Process Chemistry of Coal Utilization: Chemistry Toolkit for Furnaces and Gasifiers (Woodhead Publishing Series in Energy) [1 ed.] 0323899595, 9780323899598

Table of contents :
Process Chemistry of Coal Utilization
Copyright
Preface
Reference
Acronyms
Modeling tools and applications
Which coal quality impacts?
Terminology and prerequisites
Pulverized coal furnaces
Entrained-flow coal gasifiers
Chemical reaction mechanisms
Summary
References
Statistical prediction models
Definitions and guidelines
Using standard coal properties to interpret CQ impacts
Coals constitution and devolatilization behavior as CQ predictors
Neural nets
Summary
References
Heuristic prediction schemes
NOXLOI Predictor
Predicting CQ impacts on NOX emissions
Predicting CQ impacts on LOI emissions
Fuel grinding submodels
Char burnout submodels
Validating the LOI predictions
Novel aspects and generalizations
Comprehensive fuel quality management tools
Summary
References
Support for CFD and process simulation applications
Chemistry submodels accommodate only a handful of chemical species
Support CFD from a virtual fuels laboratory
Fuel-specific input for CFD
Thermophysical properties and drying rates
Primary devolatilization rates
Volatiles conversion
Finite-rate tar decomposition and soot production
Oxidation of char and soot
Gasification of char and soot
An important omission
Specifying representative operating conditions
Performance
Summary
References
Simulations with detailed chemistry
Laminar flamelet models vs chemical reactor networks
Historical development of chemical reactor networks
ChemNet CFD postprocessing
Overview
Implementation protocol
Delineating regions
Residence time distributions
Operating conditions
Required variables for ChemNet postprocessing
Chemical reaction mechanisms
Fully validated mechanisms
Surface reaction mechanism for soot chemistry
Soot chemistry submechanism
Implementation of soot chemistry
Species conservation equations
Equations for nonreactive carrier gas streams
Equations for gas streams with O2
Equations for gas streams with steam, CO2, H2, and CO
Summary
References
ChemNet furnace applications
ChemNet validation with data from large-scale systems
Target variables for ChemNet simulations
Case studies on PC combustion
PC combustion at lab-scale
1D coal flame structure at atmospheric pressure
Coal flame structure at elevated pressures
PC combustion at pilot-scale
Reactor network for the CRF furnace
Pilot-scale flame structure with coal
Pilot-scale flame structures with biomass cofiring
NOX predictions for pilot-scale flames with biomass cofiring
Outlook for ChemNet simulations of pilot-scale coal flames
PC combustion at commercial-scale
The impact of heat absorption on furnace NOX levels
Paths to VOC and PAH emissions
The chemical structure of a 530MWT-fired furnace
Bulk flow patterns and chemically distinct regions
Equivalent reactor network
Near-injector flame structures
Commercial CFBC furnaces
Splash zone, riser, and exit zone
Interpreting LOI emissions from commercial CFBCs
Coal quality impacts
Summary of ChemNet for furnaces
References
ChemNet gasifier applications
ChemNet validations with data from gasifiers
Target variables for ChemNet simulations
Case studies on entrained-flow coal gasification
CH4/coal co-gasification at lab-scale
Interpretations of lab-scale gasifier data
Extrapolations to commercial processing
Coal gasification at pilot-scale
Test conditions
Equivalent reactor network
Interpretations of coal conversion and syngas composition
Summary of ChemNet gasifier simulations at pilot-scale
Coal gasification at demonstration-scale
Syngas compositions
Quasiglobal homogeneous reforming mechanism
Summary of coal gasification at demonstration-scale
Coal gasification at commercial-scale
Summary of ChemNet simulations for gasifiers
References
Index
A
B
C
D
E
F
G
H
I
K
L
M
N
O
P
Q
R
S
T
U
V

Citation preview

Process Chemistry of Coal Utilization

Process Chemistry of Coal Utilization Chemistry Toolkit for Furnaces and Gasifiers Stephen Niksa Ph.D.

An imprint of Elsevier

Woodhead Publishing is an imprint of Elsevier The Officers’ Mess Business Centre, Royston Road, Duxford, CB22 4QH, United Kingdom 50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States The Boulevard, Langford Lane, Kidlington, OX5 1GB, United Kingdom Copyright © 2022 Elsevier Ltd. All rights reserved No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Details on how to seek permission, further information about the Publisher’s permissions policies and our arrangements with organizations such as the Copyright Clearance Center and the Copyright Licensing Agency, can be found at our website: www. elsevier.com/permissions. This book and the individual contributions contained in it are protected under copyright by the Publisher (other than as may be noted herein). Notices Knowledge and best practice in this field are constantly changing. As new research and experience broaden our understanding, changes in research methods, professional practices, or medical treatment may become necessary. Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any information, methods, compounds, or experiments described herein. In using such information or methods they should be mindful of their own safety and the safety of others, including parties for whom they have a professional responsibility. To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in the material herein. Library of Congress Cataloging-in-Publication Data A catalog record for this book is available from the Library of Congress British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library ISBN: 978-0-323-89959-8 (print) ISBN: 978-0-323-90381-3 (online) For information on all Woodhead publications visit our website at https://www.elsevier.com/books-and-journals

Publisher: Charlotte Cockle Acquisitions Editor: Maria Convey Editorial Project Manager: Aleksandra Packowska Production Project Manager: Debasish Ghosh Cover Designer: Matthew Limbert Typeset by STRAIVE, India

Preface

My 45-year career in coal research moved my attention from the coal pile to the smokestack at coal-fired power plants, and every unit operation in between. It also covered a host of utilization technologies considered “advanced” in their time, including atmospheric fluidized bed combustor (AFBC), pressurized fluidized bed combustor (PFBC), circulating fluidized bed combustor (CFBC), and gasifiers of varied designs. As long as the technology was either suspension fired or fluidized with coal, I was all too happy to design lab tests and develop mathematical models and simulation tools to characterize it in detail. My strongest focus has been on the chemical reaction mechanisms within furnaces and gasifiers, including formation mechanisms for an assortment of emissions (NOX, unburned carbon, soot, polynuclear aromatic hydrocarbons (PAH), volatile organic compounds (VOCs, SO3) at the furnace or gasifier exit. Another has been on minor and trace element transformations, including halogens, alkali and alkaline earth elements, and Hg, Se, and other trace metals. The genesis of these species spans the primary flame zone as well as the units in flue gas cleaning systems, over a temperature drop of 1000°C or more. My third focus has been on the chemistry of gas cleaning systems, especially the multipollutant performance of selective catalytic reductions (SCRs) on NOX, SO3, and Hg0; the deactivation of SCR catalysts by inorganic poisons; VOC transformations; and the fate of Hg species during wet flue gas desulfurization (FGD) scrubbing. During my first 20 years, I performed basic coal research as a research assistant in chemical engineering at Princeton; a staff member at Sandia National Laboratories, Livermore; and an assistant professor of mechanical engineering at Stanford. At face value, all the work from this period aimed to develop and validate chemical reaction mechanisms for coal utilization technologies. It covered coal devolatilization, secondary volatiles chemistry, and char conversion in furnaces and gasifiers. However, this work was explicitly divorced from practical applications because it only considered the behavior of isolated, individual coal particles. The opening paragraphs in my research proposals drew direct connections to the behavior of a host of utilization technologies, but the research schedules never included heavily loaded coal suspensions or any fluidized beds. And since the work was largely detached from practical applications, the findings received little attention beyond the academic research community. Notwithstanding, I was advancing a clear and important mission: to unravel the coal quality impacts in all stages of coal conversion chemistry. The goal was to advance beyond the state of understanding at the time, whereby every third or fourth coal sample was regarded as an “outlier” or “unusual coal” because its behavior was distinctive. Ultimately, I aimed to demonstrate that reaction mechanisms could accurately predict the distinctive behavior of individual coal samples, which now they can do.

viii

Preface

During the next 25 years, I worked as the Energy Program Manager at SRI International and the President of Niksa Energy Associates LLC (NEA), a small consulting firm. The imperative that matters in these commercial settings is to solve actual problems in the operations or design of coal utilization technologies. There would be no repeat business whenever a project failed to deliver a useful solution. I drew heavily on my earlier experience to specialize in the fuel quality impacts: i.e., what happens to the performance or emissions from a coal process when its typical coal is replaced by other diverse coals, or coal blends, or blends of coal with biomass or pet coke? We relied on modeling and simulations to interpret lab- and field-test data from our commercial and government clients, and kept expanding the reaction mechanisms and simulation tools to contend with the complexities of actual technologies. Early on we confronted a huge chasm in the technical information available on solutions to practical problems in coal processing. Some basic researchers can’t say enough about every facet of their work, whereas technology developers and consulting firms hardly disclose anything of a technical nature about the solutions to their clients’ operational problems. Another difference with the basic research community is that not every issue presented to a technical consultant warrants a description at the molecular level. Constraints on schedule and budget might only warrant statistical interpretations of limited performance data, whereas state-of-the-art treatments could be the only means to address other problems. Yet another difference was particular to NEA, which regarded Japan as heaven because each power plant burns a few dozen different coals from the global trade every year, so our focus on the fuel quality impacts was already ingrained. Another cultural novelty was the standing of technical consultants in Japan: In the United States, a job well done earned the label “leading expert” for that particular technical issue, and similar work was soon to follow. In Japan, a successful job earned the title “smart fellow,” so what followed were more difficult technical challenges in areas completely unrelated to the original assignment. This attitude was a huge benefit to a nascent consultancy, because acquiring new skills is an essential prerequisite for continued viability in this arena. The first volume in this series (Niksa, 2020) covers the comprehensive reaction mechanisms that were largely developed during my first 20 years, and expanded and honed for practical applications over the next 25 years. It is a useful companion to the present volume because it describes detailed operating conditions for major coal utilization technologies, including the furnaces and gasifiers in the case studies in this book. It characterizes all stages in the chemistry of coal combustion and coal gasification, including primary devolatilization, tar decomposition, tar hydroconversion, volatiles combustion and reforming, soot conversion, and char oxidation, gasification, and hydrogasification. The first volume is also loaded with datasets from lab-scale on the coal quality impacts on all stages of coal conversion. Reading the first volume is not strictly necessary to comprehend the applications in this book, but those readers will be at a marked advantage in the vocabulary used throughout this volume. As stated in the preface to the first volume, the goal of the pair of volumes is to relate the major advances in coal science to quantitative interpretations of performance data from lab-, pilot-, and commercial-scale for the major coal utilization technologies. The first volume covers coal science and this volume covers the quantitative

Preface

ix

interpretations of performance data. These interpretations are addressed with a broad variety of modeling methods—a toolkit—because “optimal” solutions to real-world problems are chosen to satisfy constraints on schedule and budget; usually, deep technical sophistication is something to avoid whenever possible. The chapters move from statistical regressions and neural nets to heuristic models to support for computational fluid dynamics (CFD) to process simulations with truly comprehensive reaction mechanisms. Each method is developed from case studies based on actual test data; in fact, all the data in this volume are authentic albeit for operating practices in commercial units over the past 30 years or so. Even though operating practices may have changed over this period, the test results are still generally representative of the technologies under consideration. The specific focus is on applications where the fuel quality impacts are first-order important. In many of these situations, the aim is to identify bad actors, i.e., the relatively few coals that are problematic among the bulk of fuels from a particular production region. Problems caused by bad actors include burner deposits, the potential for autoignition, excessive unburned carbon emissions in fly ash (loss-on-ignition or LOI), and SCR catalyst deactivation by chemical poisons, among many others. Fuel switching is another common application that involves closely managing the fuel quality impacts. It can exploit price differentials when a company switches from coal to cheaper opportunity fuels without suffering penalties of furnace derating or elevated emissions. It can also be a cost-effective route to compliance with both emissions regulations and renewable portfolio standards. In general, utility companies want to know in advance which fuels will likely minimize operational problems, and which will enhance the performance of each specific technology in their generation systems. Since chemistry largely determines the fuel quality impacts, the process chemistry figures prominently in all the case studies. The cases usually describe which reaction mechanisms come into play in each application but detailed treatments of the mechanisms are relegated to the first volume. Of course, essentially all applications at pilotand commercial-scale also involve a coupling among transport phenomena and the chemical reaction mechanisms, including bulk mixing patterns, turbulent transport, and radiation transfer. So fuel quality impacts determine the scope of the applications of interest and chemistry is the primary underlying science. And all the methods and nearly all cases in this volume pertain to the complex reacting flow fields in commercial utilization technology. The emphasis on the chemistry of fuel quality impacts excludes many important areas of practical consequence, including fuel handling and grinding; momentumdriven phenomena, such as wall impingement of coal flames; and heat transfer considerations. In addition, issues associated with mineral fouling and slagging, such as gas intercooler plugging in integrated gasification combined cycle (IGCC) systems, will not be considered. These are certainly among the most important fuel quality impacts, but they are governed by completely different physicochemical principles than the ones used here. Almost every case study in this volume utilizes PC Coal Lab, a commercial software package developed, licensed, and updated by Niksa Energy Associates LLC (NEA)

x

Preface

that predicts the rates and product distributions for all the stages of chemistry associated with coal-fired furnaces and gasifiers, except for homogeneous chemistry among species in the gas phase (for which validated elementary reaction mechanisms are widely available). Here this package is used to evaluate more powerful regression variables in the statistical methods in Chapter 2. It is the virtual fuels laboratory that predicts a fuel’s volatiles yield, nitrogen release, and char burnout history in the heuristic models in Chapter 3, and specifies all reaction rate parameters in the global rate expressions deployed in CFD simulations in Chapter 4. In the remaining chapters, PC Coal Lab is built into larger simulations of entire flow fields throughout furnaces and gasifiers. It describes the species released from the coal phase and the subsequent decomposition of tar, first into PAH and ultimately into soot, as well as the succeeding conversion of char and soot via combustion or gasification. In principle, alternative reaction mechanisms could be swapped into the same calculation sequences. But unless the alternative mechanisms had already been validated like those in PC Coal Lab (as demonstrated in Niksa, 2020), there would be no assurance that the performance would be comparable. It is a pleasure to acknowledge the very thorough proofreading of text by Dr. Brian C. Young, formerly of Envirosafe and CSIRO in Australia. During our first collaboration some 30 years ago, Dr. Young redefined for me the meaning of a publication in “final form.” His extensive edits and scrupulous cross-checking of the text in this book and the earlier volume are very much appreciated and will benefit every reader of both volumes in this series. This volume is dedicated to my main technical collaborators over the years, including Russell and Saville (deceased) at Princeton; Kerstein and Mitchell at SNLL; Chen, Marlow, Lau, Cho, Yu, and Boehman at Stanford; Eckstrom and Malhotra at SRI International; and Hurt, Kornfeld, Liu, Krishnakumar, Padak, and Naik at NEA. Thanks so much for taking on the challenges of this work with imagination, dedication, and tireless energy. The progress we made together is a shining reflection on all of us.

Reference Niksa S. Process chemistry of coal utilization: impacts of coal quality and operating conditions. London: Woodhead Publishing, Elsevier; 2020, ISBN:978-0-12-818713-5.

Acronyms

AAEM AFBC AR BO BTX CBK CCOFA CEN CFBC CFD CFS CGTL CQ CR CRF CSTR C2SM DAEM daf dmmf DPM DNS DSC EF EFR ESP ERZ FC FF FGD FNO FTS GEPS GHC GUI HGI HHV hv

alkaline and alkaline earth metals in coal atmospheric fluidized bed combustor as-received wt.% for moisture, ash, volatile matter, and fixed carbon in coal char burnout benzene, toluene, and xylene carbon burnout kinetics model as CBK/E for combustion; CBK/G for gasification by steam and CO2; and CBK/H for hydrogasification close-coupled overfire air injection CanmetENERGY, a Canadian national laboratory circulating fluidized bed combustor computational fluid dynamics Configurable Fireside Simulator for case studies on a full CFD simulation co-gasification of natural gas and coal to liquids coal quality flame core region in a ChemNet analysis of a burning coal jet Combustion Research Facility at Southern Research Institute continuously stirred tank reactor competing two-step chemical reaction model for devolatilization distributed activation energy reaction model for devolatilization dry-ash-free basis for coal composition dry-mineral matter-free basis for coal composition discrete particle submodel in CFD direct numerical simulation of a turbulent flowfield distributed control system in a power plant entrained-flow type of gasifier entrained-flow reactor electrostatic precipitator external recirculation zone in a turbulent flowfield fixed carbon content of coal fabric filter flue gas desulfurization unit for SOX control fraction of char-N converted into NO during burnout Fisher-Tropsch synthesis General Electric Power Systems light (C1–C4) gaseous hydrocarbons graphical user interface in software Hardgrove grindability index higher heating value, MJ/kg high volatile bituminous coal

xii

IRZ LES LNB LOI lv ML Mn MWD mv MW MWth NBFZ NEA NMR NSC NOX ODE OEM OFA PAH pc PCD p.f. PFBC PFR PRB p-RCFR PSD PSDF PVM RCFR REI RPM RTD SCR SDA SFOR SNCR SNOR SOFA SOX SR SSE UBC UDF VFL VOC WGSR

Acronyms

internal recirculation zone large eddy simulation of a turbulent flowfield low-NOX burner flyash loss on ignition, wt.% low volatile bituminous coal mixing layer region around a burning coal jet in a ChemNet analysis number-average molecular weight molecular weight distribution medium volatile bituminous coal furnace rating based on electricity production furnace rating based on heat release near-burner flame zone Niksa Energy Associates LLC nuclear magnetic resonance spectroscopy Nagle-Strickland-Constable kinetics for soot oxidation oxides of nitrogen ordinary differential equations original equipment manufacturer overfire air injection into a furnace polynuclear aromatic hydrocarbons pulverized coal particle collection device pulverized fuel size grade with 70 wt.% through 200 mesh pressurized fluidized bed combustor plug flow reactor Powder River Basin in Wyoming, USA pressurized radiant coal flow reactor particle size distribution Power Systems Development Facility operated by Southern Company Services proximate volatile matter content of coal radiant coal flow reactor Reaction Engineering International, a commercial CFD firm random pore model for char gasification residence time distribution selective catalytic NOX reduction spray drier absorber for SOX control single, first-order reaction for devolatilization selective noncatalytic NOX reduction single, nth-order reaction for char conversion separated overfire air injection oxides of sulfur stoichiometric ratio evaluated as the ratio of air to fuel flowrates normalized by the ratio for stoichiometric combustion standard error of estimation unburned carbon in flyash user-defined function or subroutine in CFD virtual fuels laboratory volatile organic compound water-gas shift reaction, CO + H2O $ CO2 + H2

Modeling tools and applications

1

This book is devoted to mathematical analyses for the coal quality (CQ) impacts in coal-fired power plants and gasification systems. In the broadest sense, CQ impacts comprise any aspect of a utilization technology that changes whenever coal is switched to another coal or a blend of coal with another coal, biomass, or petroleum (pet) coke. The book presents a variety of mathematical approaches—a toolkit—that can usefully interpret CQ impacts within a broad range of constraints on schedule and budget. The pace of commercial development is usually too fast for deep contemplation over the perfect method to resolve an engineering issue. And not every issue demands a detailed interpretation at the molecular level. So technical professionals need a range of options to consider before they can line up their mathematical support with an overarching project strategy. Hopefully, this book spans the gamut of approaches that can address their current and future needs. Statistical methods, particularly, multivariable regressions and neural nets are covered in Chapter 2. These methods are the simplest to implement but perform in ways that are impossible to foresee. One simply proposes some predictor variables, performs the regression calculations on a database, and then decides whether or not the statistical measures of uncertainty are within useful quantitative tolerances. Statistical methods are best suited to situations where relatively few coals are “bad actors” that cause operational problems or violate emissions regulations, and also when field test data are available on an on-going basis. The heuristic approaches in Chapter 3 combine statistical methodology with an understanding of the process chemistry, albeit at a phenomenological level. The mathematical relations to interpret the CQ impacts are made “smarter” by incorporating firmly established relations among furnace operating conditions and performance. For example, NOX emissions are inversely proportional to the proportion of air that is injected above the coal burners in a furnace, which is called the staging level. A heuristic analysis for the CQ impacts on furnace NOX emissions would incorporate this relation, along with analogous relations for many other ways that various operating conditions affect NOX production. Heuristic approaches are especially useful when the goal is to use historical field test data to screen the performance of many different prospective coals in a particular power plant or gasification facility. The chemistry submodels in computational fluid dynamics (CFD) simulations for furnaces and gasifiers are covered in Chapter 4. In many ways, CFD is a logical extension of heuristic approaches. Although the geometric fidelity is very sophisticated, the chemical reaction mechanisms are rudimentary to conform to severe constraints on computational expediency. Even so, CFD certainly relates all the most important operating conditions to the emissions and conversion efficiencies of interest. It can also resolve the CQ impacts, provided that the parameters in the chemistry submodel are specified to recreate realistic process chemistry. It is this aspect that is emphasized in Chapter 4.

Process Chemistry of Coal Utilization. https://doi.org/10.1016/B978-0-323-89959-8.00001-X Copyright © 2022 Elsevier Ltd. All rights reserved.

2

Process Chemistry of Coal Utilization

The final three chapters are devoted to simulations that incorporate the most comprehensive and fully validated chemical reaction mechanisms for the process chemistry of coal utilization. CFD cannot currently accommodate full chemistry for reacting coal suspensions, and probably will not for the foreseeable future. The simulation methods covered in this book radically simplify the flow fields and omit turbulence entirely to enable realistic process chemistry. Obviously, this skewed compromise is only suitable for certain technical issues and, fortunately, CQ impacts are among the most suitable. Chapter 5 presents the methodology, and Chapters 6 and 7 cover case studies for furnaces and gasifiers, respectively. The entire book except for Chapter 5 is oriented around case studies. They include measurements at lab-, pilot-, and commercial-scale, although utilization systems at pilot- and commercial-scale are highlighted throughout. All the data were actually recorded and none are hypothetical. So the case studies genuinely reflect the CQ impacts on the performance of utilization technologies over the past 25 years in the United States and Japan. They may not necessarily reflect current practice, but they are nonetheless realistic, particularly in the rendition of CQ impacts. The remainder of this chapter surveys which aspects of CQ are covered or omitted. Then, it briefly reviews the essential vocabulary and points to the first volume in this series (Niksa, 2020) for relevant information on the process chemistry.

1.1

Which coal quality impacts?

To anyone who has ever looked closely at a piece of coal, it is obviously not a homogeneous solid. Chunks of coal are striated into irregular layers around particles and bands of mineral inclusions. The most basic resolution of this structure has only two primary components, an organic matrix that holds particulate minerals. The organic component is actually a matrix of macromolecules mostly composed of carbon, hydrogen, and oxygen with smaller amounts of nitrogen and organic sulfur. The inorganic minerals comprise an assortment of clays, silica, alumina, calcite, and either pyrite or siderite; minor amounts of phosphorus, alkali, and alkaline earth metals; and trace levels of dozens of other metallic species, including Hg, As, Pb, and B. Almost all the major minerals appear as inherent inclusions and extraneous particles. The inherent mineral inclusions vary in size from fractions of a micron to several centimeters. When coal is ground for pulverized coal (pc) furnaces, larger inclusions are freed from the coal matrix as extraneous particles. Most of the minor inorganics like alkali and alkaline earth species are “associated” with the organic matrix through weak ionic and dispersion forces to polar functional groups in the organic matrix, rather than covalently bonded. The trace metals can appear as substitution elements in major and minor minerals, and in associations with organics. Every coal utilization technology heats the coal above its decomposition temperature when it expels a mixture of gases collectively called “volatiles” in a chemical process called “devolatilization.” The solid residue is “char” whose organic composition is depleted in hydrogen, oxygen, and sulfur, and whose inorganics have partially decomposed. Furnaces and gasifiers are designed to stage devolatilization, and also to

Modeling tools and applications

3

convert all the residual char into gaseous products. In furnaces, the primary gaseous products of char conversion are CO2 and steam, and in gasifiers, the primary products are CO and H2, whose mixtures are called “synthesis gas” or “syngas.” Once the char has been converted into gases, the inorganics in coal produce two forms of mineral products. Some gasifiers and furnaces are designed to collect minerals as molten streams called “slags” which drain from the walls into collection vessels at the bottoms of reactors or furnaces. Others recover the minerals as fly ash, which are 10–20 μm spherical mineral particles solidified from molten droplets. Regardless of the system design, flyash always forms in coal flames and will deposit onto heat transfer surfaces in furnaces as long as it remains molten, causing so-called, “fouling and slagging” problems. Once it solidifies, flyash is recovered well downstream of furnaces and gasifiers in specialized particle collection devices (PCDs). Although succinct, this introduction identifies several of the most important CQ impacts. The total volatiles yield is crucial because it indicates which portion of the coal ignites and stabilizes coal flames in a few hundred milliseconds and which remains to be converted as char over a few seconds. Moreover, volatiles yields can approach 70 wt.% with the youngest coals but are only a few percent with anthracites. Technologies designed for complete coal conversion never achieve it in practice. Since char conversion rates among different coals can be as different as volatiles yields, extents of coal conversion are another essential CQ impact. In the industry, the conversion is specified as emissions of either unburned carbon (UBC) or losson-ignition (LOI). The former term uses “carbon” as a synonym for the combustibles in coal, and the latter is the weight percentage of combustibles in flyash at the PCD. All gaseous emissions have legitimate CQ impacts, including NOX, SOX, CO, gaseous hydrocarbons (GHCs), volatile organic compounds (VOCs), and polynuclear aromatic hydrocarbons (PAH). The emissions of trace metals that also form vapors such as Hg and Se, and to a lesser extent, As and B also have CQ impacts. However, the reason that the CQ impacts on these particular emissions are emphasized in this book is somewhat arbitrary. They are chosen based on the availability of comprehensive, validated chemical reaction mechanisms to describe their genesis in furnaces and gasifiers. The first volume in this series (Niksa, 2020) presents the author’s reaction mechanisms for drying, devolatilization, tar decomposition, secondary volatiles combustion and reforming, and conversion of char and soot via both combustion and gasification. This suite of mechanisms covers all the essential transformations within the organic coal components, as well as the coal-specific aspects of volatiles chemistry once they leave the coal suspension, particularly tar decomposition into gases and PAH at moderate temperatures, and into gases, oils, and soot at elevated temperatures. When combined with validated elementary reaction mechanisms for generic chemistry in the gas phase among GHCs and light organics, these mechanisms accurately depict the production of all pollutants derived from the organic coal components. Suitable supplemental reaction mechanisms have also been developed for Hg, Se, and As, whereas mechanisms for B transformations are still under development. Conversely, the transformations among coal minerals are excluded from further consideration. This is certainly not because they do not display important CQ impacts.

4

Process Chemistry of Coal Utilization

Ash fouling and slagging behavior strongly vary among coals, and sometimes exhibit synergy among the components in coal blends. But the chemistry underlying these transformations occurs on time scales that are much longer than the transit times through furnaces and gasifiers as mineral deposits often develop over hours or even days. This makes it impossible to directly relate mineral deposition rates to the composition of the coal being fired because coal composition always fluctuates in time at power plants and gasification sites. Moreover, the mineral transformation chemistry acts in concert with mineral thermochemistry, phase changes, and specialized transport phenomena such as aerosol impaction, droplet coalescence, and surface wetting. Consequently, the CQ impacts on fouling and slagging in a particular furnace can only be sorted out after the physical phenomena have been adequately simulated. In and of itself, this prerequisite is a substantial undertaking that depends as much on the particular flowfield as on any general features of the process chemistry. This makes it difficult, if not impossible, to generalize field test data from many furnaces into a database that represents a given coal firing configuration over a sizeable furnace population. Heuristic approaches that depict the CQ impacts on fouling and slagging with reasonable accuracy are available (Erickson et al., 1995; Ma et al., 2007; Harding and Cooper, 2016), but are regarded as beyond the scope of this book. The CQ impacts on coals’ milling performance are strong and are usually gauged by the Hardgrove Grindability Index (HGI). But this relation is only considered in passing in Chapter 3. The CQ impacts on the unit operations in utility gas cleaning systems are also omitted, including economizer fouling, and the performance of selective catalytic NOX reduction units (SCRs); electrostatic precipitators (ESPs) and fabric filters (FFs); and flue gas desulfurization (FGDs) units and spray drier absorbers (SDAs). One cannot possibly comprehend CQ impacts without understanding the coal rank spectrum. First and foremost, rank denotes a sample’s geological age. Age is significant because it directly connects with the process of maturation, which denotes the biological and physicochemical processes that converted plant matter and fossil remnants, first, into peat and, ultimately, into the different types of coal. Maturation preferentially eliminated oxygen and hydrogen, which shifted the molecular components away from aliphatics toward planar condensed polyaromatic structures. As the planar aromatics grew, portions of adjacent macromolecules began to align into layers, which eventually expanded into nanocrystals within an amorphous glassy matrix. Further maturation expanded the crystalline portion at the expense of the glassy matrix. Once all the oxygen and most other heteroatoms had been eliminated, the planar aromatics formed lamella that stacked into crystallites in the macroscopic size range, as most of the glassy matrix had been eliminated. The ultimate product of maturation is not the anisotropic, purely crystalline state of graphite, whose formation requires much more severe thermal processing than could be applied to a coal deposit. Rather, it is a disordered polycrystalline form of essentially pure carbon. Maturation is the underlying, unifying principle for the coal rank spectrum because it generates only one ultimate product: that disordered, polycrystalline, carbonaceous solid. Rank denotes a sample’s geological age because age roughly denotes the extent of maturation. Of course, no one ever attempts to assign the geological ages of coal

Modeling tools and applications

5

samples to determine their ranks; many different so-called “rank parameters” are measured under standardized test conditions. The rank spectrum progresses from samples with the lowest calorific values; through progressively greater levels of fixed carbon and calorific values (from samples with progressively less oxygen), and then through progressively greater levels of fixed carbon and lower calorific values (from samples with progressively less hydrogen). The youngest coals are lignites, which are called brown coals in Australia and Eastern Europe. Low-rank coals comprise lignites and subbituminous. All grades of bituminous coals are distinguished from subbituminous and anthracites because the former melt under standardized processing conditions (like those in coke ovens), whereas coals of lower and higher rank do not. High volatile B (hvB) bituminous is lower in rank that hvA, and hvC is lower than hvB, based on calorific values. All ranks higher than hvA are collectively called low volatility coals because their volatiles yields are relatively very low. Medium volatile (mv) and low volatile (lv) bituminous coals retain the distinctive melting behavior that none of the anthracites have. These and many other aspects of CQ and coal rank are discussed further in Chapter 2 of the first volume of this series (Niksa, 2020). Throughout this book, coal rank is tracked with the percentage by mass of the carbon in a coal sample on a dry, ash-free (daf) basis. This value is normally evaluated from an ultimate analysis, which determines the mass percentages of C, H, O, N, and S in the combustible portion excluding moisture. It uses three separate tests described in ASTM D3176, plus the ash determination from a proximate analysis. A proximate analysis comprises the mass percentages of moisture, ash, volatile matter, and fixed carbon, which are obtained from a series of three standardized tests. Readers interested in more detail should consult ASTM D3173, D3174, and D3175.

1.2

Terminology and prerequisites

This book requires two forms of essential background knowledge: (1) the vocabulary used to describe the most important features and variations in coal utilization technologies; and (2) familiarity with the chemical reaction mechanisms that are ultimately responsible for the CQ impacts. The first volume in this series (Niksa, 2020) covers the first prerequisite in Chapter 1 and the second prerequisite across most of the higher chapters. Detailed descriptions of the reaction mechanisms for char conversion are also available for combustion (Niksa et al., 2003) and gasification (Liu and Niksa, 2004). Readers unfamiliar with this material are at a marked disadvantage because this brief section provides little more than the relevant vocabulary. Two coal utilization technologies are emphasized in the case studies in succeeding chapters, pc furnaces and entrained flow gasifiers. They are the most widely deployed technologies, by far, and essentially all their operating principles are rooted in firm technical foundations. Thorough descriptions of these technologies are available in reference books such as “Steam: Its Generation and Use” (Tomei, 2015), and in the extensive catalog of monographs published by IEA Coal Research (www.ieacoal.org). Here, the discussion focuses on the terminology and the different operating domains.

6

Process Chemistry of Coal Utilization

1.2.1 Pulverized coal furnaces PC furnaces process the finest size grade of coal, by far, among all coal-fired furnaces, and contact the coal with several air streams injected at various furnace elevations through convective, two-phase mixing. Gas temperatures vary by several hundred degrees from the maximum of about 1700°C in near-burner regions to the furnace exit, where temperatures are almost always 1000–1050°C. But the profile within a particular furnace is determined by burner design, the partitioning of air among burners and ports in the upper elevations, and coal quality, among other factors. Entrained-flow (EF) gasifiers are the hottest coal conversion units of all, particularly when they use dense suspensions in O2 of somewhat finer grinds than used in pc furnaces. But when EF gasifiers are fed by coal slurries, the grinds must be much coarser to manage the slurry viscosity within tolerable limits. The layout of a typical pc furnace appears in Fig. 1.1. Streams of coarse coal are ground into the pulverized fuel (p.f.) size grade in grinding mills, then conveyed in primary air streams into a manifold of burners near the base of the furnace walls. Collectively, the manifold of burners is called the burner belt, and the furnace section that

Fig. 1.1 Layout for a typical pc furnace.

Modeling tools and applications

7

contains it is called the near-burner flame zone (NBFZ). The coal stream ignites upon injection, which stabilizes a large flame across and above the NBFZ. Larger, denser particles of mostly mineral matter fall downward and are recovered as bottom ash in a hopper at the bottom of the furnace. Nearly all the coal stream moves upward while the flame mixes with so-called secondary air injected either through annular openings in the burners called registers or through separate air injection ports. Additional air is usually injected well above the burners through ports called overfire air (OFA) injectors. Beyond the NBFZ, the suspension continues upward until it is turned and accelerated through the converging superheater section. Multiple heat exchangers within the superheater section are collectively called the convective passes, based on the primary heat transfer mechanism into steam tubes in this section. The stream then passes through the furnace exit and turns downward through the reheat section and economizer. The furnace exit is the interface between the furnace and the gas cleaning system, at the superheater exit. Temperatures of 1000–1050°C into the furnace exit are too cool for appreciable coal conversion. So the gas cleaning system is not relevant to the conversion mechanisms, although it is obviously important for emissions control and the system thermal efficiency. The notable exceptions are the PCDs that recover fly ash containing UBC. Within the generic layout in Fig. 1.1, there are several variations. Firing configuration denotes the type and layout of burners or fuel injectors on the furnace walls, and there are three popular configurations in pc furnaces: Front wall-fired, opposed wall-fired, and tangential-fired. In wall-fired furnaces, either one or two burner belts contain an array of burners staggered in several rows across most of a wall. Furnaces with one belt are front wall-fired, and those with two on opposite walls are opposed wall-fired. In both variations, the streams from all the burners coalesce into one giant flame structure that bends upward before it reaches any of the other walls in the near-burner region. Tangential-fired (or T-fired or cornerfired) furnaces have fuel injectors near each corner, and several stacks of injector sets collectively called registers at multiple elevations. Multiple air ports alternate with the fuel injectors at different elevations to introduce auxiliary air and closecoupled OFA (CCOFA). Additional OFA ports above the fuel registers are called separated OFA (SOFA). The fuel injectors are adjustable in the horizontal and vertical planes to position a swirling fireball as desired within the radiant section. In T-fired furnaces, the coal streams from different injectors penetrate through much of the swirling fireball and accumulate in the central core of the flowfield while moving upward. But the bulk of gases coalesce into helical streams that remain fairly close to the walls. Two other important operating variations pertain to the amount and distribution of the air streams injected into a furnace. Excess air is the amount over and above the airflow needed to completely convert the coal feed into ultimate combustion products. In the United States, the excess air is typically 15%, which gives 3% to 4% residual O2 in the flue gas at the economizer. Elsewhere, excess air levels can be high enough to give 6% economizer O2. Of course, the stoichiometric air requirement depends on the coal composition. It is usually expressed in terms of the stoichiometric ratio (SR),

8

Process Chemistry of Coal Utilization

which is the ratio of the actual air-to-fuel ratio normalized by the value of the ratio for stoichiometric combustion, according to the following expression: mOxidizer m SR ¼  Coal  mOxidizer mCoal Stoichiometric

(1.1)

where mi is mass flowrates of an oxidizer and coal. Excess air is defined as SR minus unity, as a percentage. For typical hv bituminous coals, an excess air level of 15% requires a total airflow about ten times the coal feed rate. This multiple diminishes to seven or so for high volatility coals (because of their substantial oxygen contents) and increases slightly for low volatility coals (which contain minimal oxygen). The other important variable associated with air is the extent of staging, which denotes the partitioning of total air among the various ports on different furnace elevations. The reference condition for staging is an idealized state of complete mixing among all the air streams and the coal feed. This state could be realized if all the air was used as primary air to convey the coal into the burners or injectors. Then, the SRvalue for the primary air would equal the SR-value for the entire furnace. This scenario never arises in practice, so SR-values based on portions of the total airflow are used to express the deviation from a premixed state. The SR of the primary air stream, SRPR, incorporates only the primary airflow; that for the near-burner region, SRNB, incorporates primary plus secondary air through burners or primary plus auxiliary air and CCOFA in T-firing; and the furnace SR incorporates all air streams into the calculation. This detailed resolution of the various air streams is crucial for aerodynamic NOX abatement strategies. Values of SRNB are rarely less than 0.80 in the United States to avoid burner belt corrosion, but are often as low as 0.70 in Japan, where mostly low-sulfur coals are burned. A more cursory index for staging simply assigns the staging level as the percentage of the air injected through OFA ports above burner belts or as SOFA in T-firing. For example, if 30% of total air was injected through OFA ports in a wall-fired furnace, then the furnace would be 30% staged. The last specification for pc furnaces is the particle size. The particle size distribution (PSD) for pc furnaces is specified as 70 wt.% through a 200 standard mesh sieve [74 μm opening], and under 0.5 wt.% through 100 mesh [149 μm]. The largesize specification, called the top size, is especially important because nearly all LOI comes from the large end of the fuel PSD. For this reason, the char conversion mechanisms should target the upper half of the PSD, from a mean size of 50 μm through a top size on 50 mesh [297 μm]. Usually, the smaller half will be completely consumed and not contribute to LOI. For conventional pc furnaces, coal heating rates range from 5  103 to 5 104°C/s; maximum gas temperatures range from 1500°C to 1800°C, and total transit times are 3 to 4 s. The pressure is roughly 0.1 MPa, and O2 partial pressures vary from near-zero to 21 kPa although, immediately after ignition O2 levels rarely exceed 5 kPa except in quench layers along the walls. As noted above, coal sizes of interest range from 50 to 300 μm because only the coarser sizes affect the combustion efficiency.

Modeling tools and applications

9

1.2.2 Entrained-flow coal gasifiers EF gasifiers inject coal through the top of a refractory-lined pressure vessel, or radially into the midsection from several opposing injectors. The coal is entrained in O2 or air, and steam is introduced through dedicated injectors with or without slurry water in the coal feedstream. Flames attach to the fuel injectors and rapidly consume the O2, mostly to burn out the volatile fuel components and finest char particles. Once the local atmosphere turns reducing, char is converted into H2 and CO, the primary syngas components. Minerals are collected as a molten slag layer that flows along the walls through the bottom of the vessel. EF gasifier manufacturers do not report detailed operating conditions. However, independent studies have reported gasifier simulations for numerous test programs (Bockelie et al., 2003; Zheng and Furimsky, 2005). Here, reported fuel properties, feedstream compositions, exit gas temperatures, and syngas compositions are used to estimate operating conditions for the EF gasifiers from Shell and General Electric Power Systems (GEPS). The GEPS gasifier operates at 4.1 MPa, and the Shell gasifier operates from 2.0 to 2.7 MPa. A typical coal PSD for the Shell gasifier has a mean size of about 40 μm, finer than the p.f. size grade (Zheng and Furimsky, 2005), but the PSD for coal slurries fed to a GEPS gasifier would have a larger mean closer to 200 μm, to manage the slurry viscosity. Shell gasifiers are operated with dry coal feeds, and little or no steam is injected downstream. Consequently, the O2/coal ratio is adjusted with different coals to regulate the H/C and O/C ratios for syngas reforming. This ratio is generally increased for coals of progressively lower volatility to maintain H/C ratios from 0.80 to 1.00 and O/C ratios from 0.90 to 1.15. GEPS gasifiers are fed with coal slurry that has 65 wt.% coal, so steam/coal ratios fall between 0.50 and 0.70. The O2/coal ratios are generally increased for coals of progressively lower volatility although, as for Shell gasifiers, the variation in this parameter for coals with C-contents over 77% is minimal. For all but the high volatility coals, operators maintain H/C from 1.85 to 2.10 and O/C from 1.54 to 1.65. The thermal histories in EF gasifiers are largely unknown because such a hostile environment provides no access to any form of diagnostics. Nominal heating rates are 6000°C/s, which is slower than the estimates for pc furnaces because the sensible enthalpies of feedstreams at elevated pressure are much greater than at atmospheric pressure, all else being the same. The respective maximum temperatures are 2400°C for Shell and 2100°C for GEPS, which are significantly hotter than flame temperatures in pc furnaces. As the overall elemental compositions into EF gasifiers are similar with all coals to regulate syngas compositions, thermal histories for different coals are similar. Nominal residence times are 2.0 and 2.8 s for the Shell and GEPS gasifiers, respectively, and are probably similar for all coals. The gasification agents in these systems are O2, H2O, and CO2, and the inhibitors are H2 and CO. Their partial pressures cover broad ranges along the primary flow path because all O2 is consumed soon after injection, and the feedstream contains little, if any, CO2, CO, or H2. The compositions entering the gasifiers, based on average values of O2/coal and steam/coal, are about 90% O2 and 10% steam for Shell gasifiers and

10

Process Chemistry of Coal Utilization

45% O2 and 55% steam for GEPS gasifiers. For these gasifiers, outlet syngas compositions can be accurately estimated as equilibrium compositions along the quench cycle. These compositions are, for Shell gasifiers, 60%–65% CO; 24%–27% H2; 2%–4% CO2; and 2%–5% H2O; and, for GEPS gasifiers, they are 33%–40% CO; 23%–28% H2; 10%–15% CO2; and 21%–24% H2O. The compositions do not necessarily sum to 100% because N2 and other minor species are also present. After the different pressures for these two EF gasifiers are taken into account, the ranges of partial pressures for the gasification agents in EF gasification are the following: 1 to 2.2 MPa O2; 0 to 1.6 MPa CO; 0 to 1 MPa H2; 0 to 0.6 MPa CO2; and 0.1 to 1 MPa H2O. The impact of O2 variations is independent of the other gasification agents because combustion chemistry is much faster than gasification chemistry whenever O2 levels exceed about 500 ppm. Even so, the ranges for the other gasification agents span an enormous operating domain. In EF gasifiers, coal heating rates range from 103 to 104°C/s; gas temperatures range from 2000°C to 2400°C, and reaction times extend to a few seconds. Pressures range from 2 to as high as 8 MPa but are usually half that or lower. Oxygen partial pressures vary from 1 to 2 MPa. Partial pressures for the gasification agents in EF gasification vary from 0 to 1.6 MPa CO; 0 to 1 MPa H2; 0 to 0.6 MPa CO2; and 0.1 to 1 MPa H2O. Such variations arise spontaneously across all EF gasifiers while the O2 and steam in the feed are consumed during char conversion; CO and H2 accumulate as products in the syngas; and the syngas composition changes to equilibrate the water-gas shift reaction (WGSR), CO+H2O $CO2 +H2. Coal sizes of interest range from p.f. grinds for dry-feed systems to coarser grinds with mean sizes of 200 μm in slurry fed systems.

1.3

Chemical reaction mechanisms

Comprehensive, validated chemical reaction mechanisms are essential for the simulations with full chemistry developed in Chapter 5 and applied in Chapters 6 and 7. They are also essential for the statistical methods in Chapter 2, to specify predictor variables with superior performance; and for the heuristics in Chapter 3, to act as virtual fuel laboratories (VFLs) for simplified flowfields that mimic much more complicated flows; and for the CFD chemistry submodels in Chapter 4, to specify the parameters in the rudimentary rate expressions in CFD that recreate the performance of comprehensive reaction mechanisms. Indeed, the solutions in nearly all the case studies in this book were developed from comprehensive reaction mechanisms, either directly or indirectly. This section surveys the phenomenological basis for the essential reaction mechanisms, primarily as a means to review the vocabulary needed to discuss their performance in detail. It does not survey how they were developed; what they include and omit; what parameters they contain; and how those parameters are specified. Similarly, it does not cover the extent of their validation databases or their coverage of the CQ impacts. Suffice to say that each mechanism has been thoroughly validated with lab-scale datasets for any coal sample across the rank spectrum for the full

Modeling tools and applications

11

domains of operating conditions in pc furnaces and EF gasifiers. Moreover, the development, performance, and parametrization of these mechanisms are thoroughly described in the first volume of this series (Niksa, 2020). In both furnaces and gasifiers, a coal’s thermal decomposition initiates a cascade of distinct chemical processes within and around the fuel particles, for two reasons. First, the surrounding entrainment gases heat at rates comparable to those for the suspended coal particles, so volatiles continue to decompose after their release from the coal. Second, process streams almost always contain reactive gases, particularly O2 or H2, that accelerate the conversion of volatiles into ultimate products. The cascade begins with “primary devolatilization” which generates volatiles from chemical processes within only the condensed coal phase. It is always the first chemical process, which may initiate some or all of the conversion sequences in Fig. 1.2. This diagram shows the major products of primary devolatilization, secondary volatiles pyrolysis, volatiles combustion, and volatiles reforming. Species preceded by “+” are produced, whereas “” denotes consumption. So-called “primary volatiles” comprise primary tars, noncondensable fuels (C1-C3 GHCs, CO, H2, HCN, H2S), and the gasification agents, steam and CO2. When neither O2 nor H2 are present in the surrounding atmosphere, “secondary volatiles pyrolysis” (a.k.a. “secondary pyrolysis” or “volatiles pyrolysis”) converts primary volatiles into secondary pyrolysis products. The predominant transformation is “tar decomposition,” which converts primary tars into secondary tars, oils, and additional noncondensables, and ultimately forms PAH, and/or soot plus CO, H2, H2S, and HCN, depending on the ambient temperature. Volatiles pyrolysis also reforms the GHCs into CH4 and C2H2 and can produce NH3. In the presence of O2, “volatiles combustion” comprises the combustion of the mixtures of noncondensable fuel components in secondary volatiles, as well as the conversion of volatile-N species such as HCN and NH3 into noxious gases and the conversion of H2S into SO2. In the presence of H2 and gasification environments, “volatiles reforming” denotes the conversion of noncondensables under reducing atmospheres, including the hydrogenation of tars and GHCs into olefins and oils and water-gas shifting. The major reaction processes in a devolatilization stage are primary devolatilization, tar decomposition, and other aspects of secondary volatiles pyrolysis and, if the atmosphere is reactive, volatiles combustion or volatiles reforming. The combustion of soot and char may compete for the available O2 with gaseous fuel compounds, depending on the operating conditions, as seen in Chapters 5–7. But soot and char gasification are relegated to a different stage because their characteristic time scales are much longer than those for any of the chemistries of the devolatilization stage. Primary devolatilization only involves heterogeneous chemistry, by definition, but all subsequent chemical conversions of volatiles are homogeneous in the gas phase, except for a few heterogeneous reactions in which soot and char catalyze chemistry among a very limited number of volatiles. For example, NO formed during volatiles combustion can be reduced into N2 by CO on soot and char. But this heterogeneous transformation should be considered an aspect of char oxidation because char oxidation is the source of the CO reductant. Also, tars may deposit onto char within a coal’s internal pore

Fig. 1.2 Chemical processes within a devolatilization stage.

Modeling tools and applications

13

system, but only when heating rates are much slower than those imposed in furnaces and gasifiers. So this heterogeneous transformation will be ignored. In the first volume of this series (Niksa, 2020), Chapters 4–6 cover primary devolatilization; Chapter 7 covers tar decomposition; Chapter 8 covers volatiles reforming and volatiles combustion; and Chapter 9 covers hydropyrolysis, tar hydrogenation, and the hydrogasification of soot and char. Collectively, these reaction mechanisms cover all the process chemistry in our subject utilization technologies, except char and soot conversion. This author’s approach to these essential heterogeneous processes is also available (Niksa et al., 2003; Liu and Niksa, 2004). Before all this chemistry is applied in the case studies, it is fair to ask, “Why resolve the devolatilization stage into a sequence of distinct chemical processes at all?” As the time scales for primary devolatilization and volatiles conversion are comparable, is there more than academic interest to be gained by all this complexity? Shouldn’t the devolatilization stage be regarded as a single chemical process? There are two very good reasons to resolve the distinct chemical stages of a devolatilization stage in practical applications. First, the coal constitution determines primary devolatilization behavior but has virtually no impact on volatiles conversion. In other words, the connections among coal constitution and devolatilization behavior are readily apparent in the distributions of pristine primary products but obscured by volatiles conversion chemistry. In fact, the knowledge that became the basis to predict the devolatilization behavior of individual coal samples was mostly revealed by measured distributions of primary products from numerous coals. The second reason is that nearly all the chemistry required to describe volatiles conversion was already elucidated by the combustion kinetics community before volatiles conversion chemistry was addressed with modern approaches. Once primary devolatilization and tar conversion are factored out, volatiles conversion chemistry is completely covered by conventional homogeneous combustion chemistry among gaseous species. The oxidation of mixtures of coal volatiles abides by the mechanisms developed for the combustion of natural gas, synthesis gas, and gas mixtures from sources that have nothing to do with coal. Coal-derived gas mixtures tend to be among the most complex but can nevertheless be analyzed with conventional oxidation mechanisms. These same generalizations pertain to volatiles pyrolysis and volatiles reforming as well. But tar conversion has not yet been analyzed with the elementary mechanisms developed for the production of soot from combustible mixtures of natural gas and heavier hydrocarbon gases, mostly because tar structures are almost as complex as coal macromolecules. Accordingly, the devolatilization stage is subdivided into distinct reaction processes as the only practical means to elucidate the connections between coal constitution and the crucial partitioning of coal into volatiles and char; and to utilize the phenomenal knowledge base on homogeneous conversion mechanisms and kinetics in simulations of coal processing.

14

1.4

Process Chemistry of Coal Utilization

Summary

Historically, CQ impacts on the performance of furnaces and gasifiers were managed with testing, sometimes at pilot-scale but mostly with field tests at commercial scale. Databases were analyzed with rudimentary statistical methods and predictor variables were drawn from a coal’s proximate and ultimate analyses, in the hope that regressions would emerge with some predictive capability. But polynomial regressions are almost never accurate enough to predict the performance of particular coal samples in utilization technologies, so they never displaced additional testing. An emphasis on taking data is not an effective CQ management strategy because it is expensive, time-consuming, and routine. This book develops an alternative strategy that regards field test data as the means to demonstrate the predictive capabilities of more sophisticated mathematical models for the CQ impacts on the target performance characteristics. These models are developed directly from chemical reaction mechanisms that have already been stringently validated under closely controlled conditions in tests at the lab-scale. Even though lab tests cannot possibly recreate the extremely complex reacting flowfields in furnaces and gasifiers, they are the best available means to validate the essential reaction mechanisms. And once validated, the reaction mechanisms are the most accurate means to resolve and predict the CQ impacts at a technological scale. After all, the basis for any CQ impact is process chemistry. Hopefully, the utility of the reaction mechanisms will become clearer as a reader progresses through cases on statistical methods, heuristics, CFD, and simulations with full chemistry in the succeeding chapters.

References Bockelie MJ, Denison MK, Chen Z, Senior CL, Sarofim AF. Using models to select operating conditions for gasifiers. In: Proc. Pittsburgh Coal Conf., Pittsburgh, PA; 2003. Erickson TA, Allan SE, McCollor DP, Hurley JP, Srinivasachar S, Kang SG, Baker JE, Morgan ME, Johnson SA, Borio R. Modelling of fouling and slagging in coal-fired utility boilers. Fuel Process Technol 1995;44:155–71. Harding NS, Cooper SA. Boiler performance and cost analysis of fuels and fuel blends using the Fuel Quality Advisor. Fuel Process Technol 2016;141:185–95. Liu G-S, Niksa S. Coal conversion submodels for design applications at elevated pressures. Part II. Char gasification. Prog Energy Combust Sci 2004;30(6):697–717. Ma Z, Iman F, Lu P, Sears R, Kong L, Rokanuzzaman AS, McCollor DP, Benson SA. A comprehensive slagging and fouling prediction tool for coal-fired boilers and its validation/ application. Fuel Process Technol 2007;88:1035–43. Niksa S. Process chemistry of coal utilization: impacts of coal quality and operating conditions. London, UK: Woodhead Publishing, Elsevier; 2020, ISBN:978-0-12-818713-5. Niksa S, Liu G-S, Hurt RH. Coal conversion submodels for design applications at elevated pressures. Part I. Devolatilization and char oxidation. Prog Energy Combust Sci 2003;29 (5):425–77. Tomei GL, editor. Steam its generation and use. 42nd ed. Charlotte, NC: Babcock and Wilcox; 2015. Zheng L, Furimsky E. Comparison of Shell, Texaco, BGL, and KRW gasifiers as part of IGCC plant computer simulations. Energy Convers Manag 2005;46(11/12):1767–79.

Statistical prediction models

Nomenclature A AC/Cl CQ E EA fa0 H 0 fa fal Fb(0) FC N fCHAR HHV0 L M MR O2 RFuel RVit S% SALIPH SC SRPR TAD VMar VMdaf W Wij W/D YGAS YGHC YOXY YTAR

ash content, as-received wt.% number of aromatic carbons in a nucleus in coal coal quality mass percentage of element E ¼ C, H, O, N, S, daf wt.% apparent activation energy for autoignition, kcal/mol carbon aromaticity of coal proton aromaticity of coal mass fraction of aliphatic material in coal fraction of linkages that are labile bridges furnace firing configuration or fixed carbon in a proximate analysis fraction of coal-N in char higher heating value of coal normalized by the value for carbon furnace load, % moisture content, as-received wt.% reactive macerals as the sum of vitrinite plus liptinite excess O2 at a furnace exit fuel ratio of fixed carbon over volatile matter vitrinite reflectance percentage of staged air in a furnace aliphatic sulfur in coal, daf wt.% staging configuration stoichiometric ratio for primary air plus coal adiabatic temperature for a volatiles/primary air flame, °C proximate volatile matter on an as-received basis, wt.% proximate volatile matter on a dry, ash-free basis, daf wt.% total volatiles yield, daf wt.% weight factor for neuron i into level j in a neural net index on wet or dry bottom furnaces yield of noncondensable gases, daf wt.% yield of chain hydrocarbons, daf wt.% yield of oxygenated gases, daf wt.% yield of tar, daf wt.%

Greek symbols τF Ψ

transit time through a furnace, s parameter defined in Eq. (2.1)

Process Chemistry of Coal Utilization. https://doi.org/10.1016/B978-0-323-89959-8.00003-3 Copyright © 2022 Elsevier Ltd. All rights reserved.

2

16

Process Chemistry of Coal Utilization

Subscripts B

pertaining to labile bridges

This chapter covers statistical models for the CQ impacts with case studies on emissions and coal conversion efficiencies. The CQ impacts are the responses in the quantity of interest to changes from one coal to another, or to a blend of multiple coals, or blends of coal and biomass or coal and pet coke. As coals exhibit continuous variations across the rank spectrum, CQ impacts should also be regarded as a continuous variation. But the question is, “A variation in which particular properties?” Statistical regressions address this question by evaluating the relation among a dependent variable—the quantity of interest to be predicted—and any number of independent variables, aka “predictors.” They are particularly easy to use because the formal analyses also provide a numerical index from zero to unity on the strength of the relation between the dependent variable and the predictors. A multitude of different regression analyses are available, and often many different forms can interpret the same database to within only minuscule differences in performance. This chapter covers only two, multivariate regressions and neural nets, simply because these are used most often to interpret performance data from coal utilization technologies. The CQ impacts are emphasized throughout, although some case studies also cover variations in the process operating conditions. In the presentation, NOX emissions and UBC from large coal flames are the common representative examples of dependent variables. Readers unfamiliar with the technology and terminology in full-scale utility furnaces pertaining to NOX and UBC should consult Chapter 1 in the first volume of this series (Niksa, 2020) for a brief tutorial, or more comprehensive treatments (Tomei, 2015).

2.1

Definitions and guidelines

Regressions are statistical relations among some quantity of interest and a database of measurements that, in some known or unknown ways, affect the values of the subject quantity. For example, NOX emissions from utility furnaces are known to depend on CQ as well as the firing configuration (FC), furnace load (L), excess O2 at the furnace exit (O2), extent and configuration of air staging (S%, SC), and whether or not the burner belt is operated for slagging or nonslagging mineral behavior (W/D). This example has one dependent variable, the NOX concentration at the smokestack, and seven predictors (CQ, FC, L, O2, S%, SC, W/D). It is best to handle variables with only a few discrete values differently than the continuous variables. For example, there are only a few firing configurations (T- or corner-fired, wall-fired, arch-fired, turbo-fired), staging configurations [low-NOx burners (LNBs), close-coupled overfire air (CCOFA), separate overfire air (SOFA)], and slagging behavior (wet or dry bottom). So it is best to subdivide a database according to the specifications on discrete predictors, and use only continuous variables as predictors for each subgrouping. Such subdivisions imply that NOX is a

Statistical prediction models

17

Fig. 2.1 (Left) Separate regressions for NOX emissions from a pilot-scale flame for three staging levels, and (right) parity plot for () coals used to develop a regression; and additional (●) coals and (▲) coal blends used to evaluate the predictive capabilities.

function of CQ, L, O2, and S% or, more formally, NOX ¼ NOX (CQ, L, O2, S%) for each of the data subsets. Regressions entail the simplest mathematics that can be used to interpret data into a predictive model, and all statistical software packages have multivariate regression routines that essentially automate the implementation. Even so, regressions are neither simple to formulate nor straightforward to assess. The most significant issues are illustrated with the regressions for NOX emissions from two pilot-scale coal flames in Fig. 2.1. One dataset covers several different coals fired with different staging levels (S% ¼ 0%, 15%, and 25%), and the other covers several world-traded coals, plus additional suites of coals and coal blends fired at nominally the same furnace conditions. The first dataset was subdivided by S%, then linear regressions were developed in the product of the coal-N content and the fuel ratio, RFuel, which is the ratio of the fixed carbon to volatile matter in a proximate analysis. The fuel ratio is the traditional predictor for CQ impacts and NOX emissions certainly depend on coal-N contents, so using their product seems reasonable. However, the regressions of measured NOX have correlation coefficients, r2, from 0.67 to 0.83, which means that this predictor explains only two-thirds to four-fifths of the total variance in the dataset. The standard deviations vary from 12 to 15 ppm, which may or may not be acceptable in a particular application. Bear in mind that these performance statistics pertain to the regressions for these three datasets, and have nothing to do with the performance of the regressions as prediction models for NOX emissions. The predictive capability is evaluated by the second dataset, which was also recorded in a pilot-scale flame with multiple coals and coal blends fired at nominally the same operating conditions. The regression under evaluation was more complex than the first case but still uses only predictors from the proximate and ultimate analyses for the CQ impacts. The parity plot shows that the regression is extremely accurate for the six coals used to specify the regression coefficients (only because the number of predictors was comparable to the number of coals in the dataset). But it is similarly accurate with only two of the nine coals used to evaluate the predictions,

18

Process Chemistry of Coal Utilization

and with none of the blends. Even if the tolerance is expanded to 50 ppm, less than half the additional coals and none of the blends qualify. These examples illustrate several important aspects of statistical regressions for CQ impacts: (i) The size of a regression database definitely matters. Whenever the numbers of data cases and predictor variables are comparable, performance statistics for the regression will be much better than those for predictions from the regression model. As the coal rank spectrum covers a very broad range of coal constitution, many coals must be tested to even see the actual CQ impacts. How many coals must be tested to cover the entire rank spectrum? This author finds that CQ impacts begin to emerge in data on a dozen coals, and that approximately two dozen coals are needed to specify a regression model whose performance statistics in predictions are similar to those in the regression analysis. In other words, in the second example in Fig. 2.1, if two dozen coals had been used to specify the regression instead of only six, then the performance statistics on the predictions for the additional coals and coal blends would have been roughly the same as for the regression analysis. Fewer coals need to be tested when only a portion of the rank spectrum is of interest. (ii) Performance statistics on a regression analysis have nothing to do with the performance of a regression model in predictions. Consequently, two independent databases are required, one for the regression analysis and one to validate the predictions from the regression model. Usually, performance statistics on the predictions are updated over time as additional test results are compiled and compared to predictions from the regression model. So the performance of any statistical prediction model may vary over time, and certain coal ranks are often seen to have more accurate predictions than other ranks. (iii) When a regression database is subdivided, the performance statistics on the regression and the predictions for each subgrouping are often quite different. In principle, the subgrouping factors out the influence of one or more variables to highlight the sensitivity to the CQ impacts. In the first example in Fig. 2.1, the impact of staging was factored out by developing separate regressions for each staging level. As the number of data cases in each subgrouping are comparable, one could reasonably expect comparable resolution of the CQ impacts. Yet the correlation coefficients for the regressions varied from 0.67 to 0.83, which is appreciable. There is no assurance that CQ impacts will be equally resolved over subgroupings in neither a regression nor in the predictions from a regression model.

Just as regressions for CQ impacts in different subgroupings often perform differently, regressions of data from different flame sizes can also perform differently. The underlying premise on each set of data is that values for a dependent variable and all the predictors have been simultaneously specified. In other words, there is a one-toone correspondence among the dependent variables and all predictors. In a lab-scale flame, coal samples are prepared in batches that are large enough to sustain numerous test runs, so the coal fed into each run is the same, and all operating conditions are monitored continuously and simultaneously. In pilot-scale flames, barrels of each coal are classified and analyzed before tests are run and, again, the coal fed into multiple runs will be very similar. The only ambiguity is that it is difficult to feed preclassified coal in a representative manner, so the grind as fed can be affected by inadvertent classification during storage and in the feeding system. There will also be time lags of a few seconds between feeding and monitoring stations, but they will not matter as long as the coal feed is truly uniform.

Statistical prediction models

19

However, in full-scale furnaces, each coal delivery may be sampled and analyzed from numerous batches, but it is still nothing like a homogeneous sample with uniform properties. Moreover, once coal lands in a stockpile, it can be conveyed to the mills with other coal deliveries, deliberately or inadvertently blended with other coals in storage, left to weather under inclement conditions, or otherwise adulterated. Mills and coal delivery manifolds classify a coal stream in unregulated ways. Under the best circumstances, coal samples are pulled from mill outlets at the beginning of each testing day despite the potential for appreciable variations in the coal properties over much shorter times. And time lags from when a coal sample is fed until its flue gas leaves a smokestack can be as long as half a minute in modern gas cleaning systems. These variations are illustrated in Fig. 2.2 with comparisons of the spread in measured values for fixed carbon and proximate volatile matter. The data labeled as “Rec0 d” used samples taken as the coal was delivered, and those labeled as “Fired” used samples taken downstream of the mills. Coal storage and handling removed the coal components with the greatest fixed carbon/least volatile matter, so that the coals actually burned were markedly more volatile than the delivered coals. Simply put, the relation between reported coal properties and the coal properties actually responsible for the gas composition monitored at a stack is fraught with ambiguity. These uncertainties can neither be characterized in any formal way nor eliminated from a statistical analysis. They should be kept in mind whenever an analyst wants to continue adding predictors to a regression to perturb a correlation coefficient Fixed Carbon 65%

Volatile Matter 45%

40% 60% 35%

55%

30%

25% 50% 20%

45%

FC Rec'd FC Fired

15%

VM Rec'd VM Fired

Fig. 2.2 Spread in measured values for (left) fixed carbon and (right) proximate volatile matter in coals to power plants as delivered (labeled as “Rec0 d”) and as fired (labeled as “Fired”). Reproduced with permission from DeAngelo JG, Sjoberg JE. The effect of coal quality on meeting the 1995 ozone season NOX cap at New York State Electric and Gas. Prog Energy Combust Sci 1999;25:341–52, Elsevier.

20

Process Chemistry of Coal Utilization

upward in small increments because fine-tuning amid such substantial uncertainties on the predictors is meaningless.

2.2

Using standard coal properties to interpret CQ impacts

CQ variables should be easy to evaluate from standard coal property data. Ideally, they can be directly evaluated from the proximate and ultimate analyses, perhaps with additions such as grind specifications. Variables based on sophisticated analytical techniques are fine, provided that the end-user is committed to compiling an extensive database of measurements for dozens of coal samples. This section presents representative regressions for CQ impacts based on standard coal properties, whereas more powerful predictors are covered in the following section. The first case study developed a prediction model for UBC emissions from an entire fleet of T-fired furnaces in India (Sathyanathan and Mohammad, 2004). UBC is evaluated as 100  XCHAR, where XCHAR is the percentage of combustibles burnout. The UBC values are not proportional to LOI because the latter quantity depends on the ash content of coal (A) as well as UBC. The furnace population for the regression database comprised 16 furnace ratings of 60 MW; 20 for 110 MW; 55 for 200 MW; and 5 for 500 MW. The analysis began with a thorough survey of the numerous factors affecting UBC emissions, including CQ factors (proximate and ultimate analyses, higher heating value (HHV), ash melting behavior, grind fineness, and top size); furnace configuration (mills-in-service, number of injector elevations, furnace residence time, τF); and staging levels (excess air, S%, SOFA, CCOFA). These parameters were then refined with lab testing and preliminary screening evaluations into a set of four predictors: RFuel, log(A), HHV0 , and τF, where HHV0 is the higher heating value normalized by the value for pure carbon (HHV/7842 kcal/kg). In addition to the individual predictors, exploratory regressions identified the following variable as especially powerful: Ψ¼

RFuel + HHV0 + log ðAÞ τ2F

(2.1)

For subgroupings based on boiler ratings of 60, 110, 200, and 500 MW, the best threevariable regressions of UBC emissions were with RFuel, log(A), and Ψ . This set of predictors gave correlation coefficients greater than 0.90 for each rating group. The most effective single predictor was Ψ . Ultimately, the rating-based regressions were collapsed into the following fourth-order polynomial in Ψ that predicted UBC emissions for all furnace ratings for the full range of Indian coals: UBC ¼ 3  106 Ψ 4 + 4  104 Ψ 3  1:61  102 Ψ 2 + 2:97  101 Ψ  0:944

(2.2)

Statistical prediction models

21

Table 2.1 Validation of predicted LOI from Eq. (2.2). LOI RFuel

HHV0

Log(A)

Ψ

Measured

Predicted

1.564 1.286 1.193 1.153

42.4 38.4 42.7 41.6

1.588 1.637 1.619 1.601

7.29 6.61 7.28 7.10

0.54 0.43 0.51 0.49

0.52 0.43 0.52 0.50

Reproduced with permission from Sathyanathan VT, Mohammad KP. Prediction of unburnt carbon in a tangentially fired boiler using Indian coals. Fuel 2004;83:2217–27, Elsevier.

For the entire furnace population, the correlation coefficient for this regression is 0.826, which is remarkable considering the numerous sources of random uncertainties in this database. Predicted UBC emissions are validated with data from several 500 MW furnaces in Table 2.1. The accuracy is especially compelling for the broad range of RFuel covered by these test conditions. It is no coincidence that this study compiled an unusually large database on commercial furnace populations.

2.3

Coals’ constitution and devolatilization behavior as CQ predictors

The main advantage of using a coal’s proximate and ultimate analyses as a source of predictors for CQ impacts is that these tests are readily available and almost universally reported with field test results. Even so, the actual quantities that are measured in the standard coal properties are too coarse to reflect even the most important aspects of coal constitution, as elaborated in Chapter 2 of the first volume in this series (Niksa, 2020). This section considers regressions based on predictors that supplement those from standard properties with either specialized analytical information that directly pertains to a subject-dependent variable, or with much clearer connections to coal constitution. In both situations, the goal is to utilize more powerful predictor variables to improve the accuracy of the prediction model. Case studies at the end of the section use predicted devolatilization yields and product characteristics as predictors. The first case study used a regression database of UBC measurements from a 1 MW pilot-scale flame for 16 world-traded coals (Cloke et al., 1997). UBC is defined as the remaining percentage of the combustibles on a dry, ash-free (daf) basis. For the regression analysis, the analysts proposed a combination of standard coal properties (C, A, RFuel) and unconventional predictors: vitrinite reflectance (RVit), reactive macerals (MR as the sum of vitrinite plus liptinite), and 190 unreactives (U190). The values for RVit and MR were obtained with automated optical microscopy according to conventional protocols (Cloke et al., 1997). The values for U190 were specified from a custom optical analysis that monitored coal reflectance across the entire greyscale in blue and white light. The examination provides an expedient and automated means to estimate the proportion of unreactive combustibles in coal,

22

Process Chemistry of Coal Utilization

analogous to an inertinite level. The underlying premise for this predictor is that combustibles in coals contain reactive components that hardly contribute to UBC (liptinite and vitrinite) as well as nonreactive components that contribute the bulk of UBC. Even if nonreactives are partially consumed during char burnout, their residual amount remains proportional to the level in the initial combustibles. Single-variable regressions were developed from the UBC database for each of the six predictors, which rank-ordered them as U190, MR, RFuel, C-Content, RVit, and A. The performance as regression variables was highly variable, and the correlation coefficients varied from 0.80 for U190 to 0.135 for A. Unfortunately, neither multivariable regressions nor evaluations of predictions with a separate validation database were reported. Notwithstanding, these regressions demonstrate that more powerful predictors for UBC emissions are available than the standard coal properties. The next two case studies use predictors determined by the coal constitution submodel in a mechanism for the primary devolatilization of any coal at any operating conditions called FLASHCHAIN. This mechanism represents coals’ crosslinked macromolecular structure as a mixture of chain fragments ranging in size from a monomer to the nominally infinite chain. The diverse assortment of structural components in real coals is rendered coarsely with four structural components: aromatic nuclei, labile bridges, char links, and peripheral groups. Aromatic nuclei are refractory units having the characteristics of the hypothetical aromatic cluster inferred from 13C NMR spectra. They also contain all the nitrogen in the coal. Except for HCN production from their nitrogen, nuclei are not transformed during pyrolysis. Nuclei are interconnected by two types of linkages, labile bridges and char links. Labile bridges represent groups of aliphatic, alicyclic, and heteroatomic functionalities, not distinct chemical bonds. They contain all the oxygen and sulfur and the aliphatic carbon and hydrogen, but no aromatic components. Peripheral groups are the remnants of broken bridges on fragment ends that have the same C/H/O/S composition. Being refractory, char links initially present in the coal are completely aromatic with no heteroatoms. But labile bridges that decompose into char links during devolatilization leave a fraction of their oxygen in the char link. This residual oxygen is released as CO at high temperatures. In the chemical reaction mechanism, labile bridges are the key reaction centers. The population of labile bridges contains the pool of all aliphatic hydrocarbon elements and all oxygen, but none of the aromatic constituents. Consequently, their elemental compositions are radically different than values for whole coals, as seen elsewhere (Niksa, 1994, 2020). The crucial implication is that the elemental compositions of whole coals are very poor indicators of the compositions of their reaction centers. Accordingly, values of atomic hydrogen-to-carbon ratios for bridges, denoted as (H/C)B, actually increase for coals of progressively higher rank, whereas wholecoal ratios from the ultimate analysis diminish. Across the rank spectrum, bridgebased values double from 1.5 to 3 while whole coal values are halved from 0.8 to 0.4. The reason is that the number of labile bridges available to incorporate the aliphatic hydrogen diminishes much faster with coal rank than the pool of aliphatic hydrogen does. The rank dependence of (O/C)B values does not go counter to that for the whole coal property. But the magnitudes of (H/C)B, (O/C)B, and (O/H)B are roughly twice the whole-coal values. The sample-to-sample variability of these

Statistical prediction models

23

Table 2.2 Bridge composition parameters from FLASHCHAIN. Bridge-based

Definition

C-number

ð1fa0 ÞAC=Cl fa0 ð1βÞ

H/C O/C N/C S/C


H ð1H fa0 Þ C 1f 0
 ð aÞ O C

1

ð1fa0 Þ

0
SALIPH  C

1

ð1fa0 Þ

atomic ratios is much larger in the bridge-based values, making them superior regression variables for devolatilization modeling and, in all likelihood, many other forms of coal processing. The atomic ratios for H/C, O/C, and S/C for bridges are evaluated with the relations in Table 2.2 from only the proton-aromaticity (Hfa0 ) and carbon-aromaticity (fa0 ), and the respective atomic ratios based on a conventional ultimate analysis of the whole coal, where the ratio for sulfur includes only aliphatic-S. As quaternary nitrogen usually constitutes a small fraction of coal-N, the bridges in FLASHCHAIN contain no nitrogen at all. Both aromaticities are strongly correlated with a coal’s C-content, and the following regressions of literature data are used for both quantities to circumvent a sophisticated and expensive laboratory analysis for every sample of interest (Niksa, 1991): fa0 ¼ 1:59  102 C  0:564

(2.3a)

H 0 fa

(2.3b)

¼ 1:10  102 C  0:554

To evaluate a C-number, the number of aromatic carbons per nucleus, AC/Cl, is also required, which is estimated from AC=Cl ¼ 0:373C  15:09

(2.3c)

The next case study uses bridge-based composition parameters as predictors for the nominal activation energy for coal autoignition. A coal’s autoignition behavior refers to its tendency to spontaneously ignite and burn under ambient conditions. It is also called “spontaneous combustion,” “spontaneous ignition,” or “self-heating.” Autoignition is a serious concern in any large-scale coal utilization operation for several reasons. It can be a serious safety hazard, and the financial penalties due to loss of calorific value can be substantial even when the combustion is very slow. Autoignition also affects a coal’s chemical and mechanical properties, especially the thermoplastic behavior that liquifies bituminous coals during thermal processing. In practice, coals

24

Process Chemistry of Coal Utilization

autoignite when stored in large volumes like holding piles and rail cars. Factors that affect air circulation and heat conduction affect the autoignition tendency, including the volume of the coal pile, its bulk density, moisture content, and extent of channeling, the particle size distribution, the relative humidity, and the wind speed. Coal rank is a predominant consideration. A coal’s potential for autoignition is often assessed in lab tests that monitor an oxidation rate while a sample is slowly heated in air. The rate data are then analyzed with an enthalpy balance to evaluate an apparent activation energy for autoignition, EA. The greater the value the stronger the potential for autoignition. This case study uses a regression database of assigned activation energies for 26 world-traded subbituminous and hv bituminous coals to develop a prediction model to replace the lab tests. Numerous predictor variables were evaluated, including standard coal properties (VMar, VMdaf, M, H/C, O/C), structural parameters from Table 2.2 ((H/C)B, (O/C)B, the aliphatic fraction, fal, the fraction of linkages that are labile bridges, Fb(0)), and various aspects of devolatilization behavior predicted for the tests that determined EA-values (weight loss (W), YTAR, YGAS, oxygenated gas yields (YOXY), and gaseous hydrocarbon yields (YGHC)). The most popular standard properties for CQ impacts on autoignition potential are VMar and M. But as seen in Fig. 2.3, they performed poorly with this regression database, and account for only 17% of the variance. Moreover, the predicted EA-values are badly biased as the predicted values are too large at low magnitudes and vice versa. Among the nontraditional predictors, the various volatiles yields did not perform as well as either the standard properties or the structural parameters. The main reason is that the maximum temperature in the test was only 250°C, which is cooler than the temperature range of the validation database for FLASHCHAIN. But a combination of standard and structural predictors gave the best regression. As seen in Fig. 2.3, this regression uses (H/C)B, O/C, VMdaf, and M as predictors. Its performance was significantly better than the best three-variable regression, but adding a fifth variable was inconsequential. This regression eliminates the bias in the regression with VMar

Fig. 2.3 Parity plots for EA-values based on (left) VMar and M and (right) (H/C)B, O/C, VMdaf, and M.

Statistical prediction models

25

and M, although the correlation coefficient is only 0.54. This low value could be indicating that measurement uncertainties were primarily responsible for the scatter; in fact, when replicate measurements were replaced with mean values for pairs and intermediate values for triplets, r2 increased to 0.70. A second validation database was not available to evaluate the best regression as a prediction model. So the database was split into Phase I and Phase II subgroups, one used as a regression database and the other as a validation database. Then these roles were reversed in a second calculation pass. Parity plots on the model predictions appear in Fig. 2.4. The regression appears to offer significant predictive capabilities, albeit in only one of the two cases. The right panel shows the better correlation of the Phase I database applied to the Phase II database. The reliability of the predictions is even better than it appears, for two reasons. First, a single point on the graph at a measured value of 10.5 kcal/mole actually indicates two superimposed tests in which the predicted and measured values are virtually identical. Second, the two cases with discrepancies of more than 1 kcal/mole are actual cases with two and three replicate measured values. In fact, the predicted values are in very close agreement with one or more of the available measured values for these two coals. The left panel in Fig. 2.4 shows the best correlation of the Phase II database applied to the Phase I data. In this case, the discrepancies have grown, but the correlation still depicts the correct tendency across the entire range of measured values. The next case study illustrates how predicted devolatilization yields can be used as predictors. It is based on new operational problems that an OEM experienced when it started using coals from a new source country. Occasionally, one of these coals would form deposits within the throats of the fuel injectors at their utility furnaces, which disrupted the staging levels and flow patterns in the burner belt regions. They characterized this tendency with a simple lab test that heated a thin layer of coal on a hot plate to different temperatures for three minutes. After heating, the plate was shaken to remove loose particles, and the deposit was weighed and assigned a numerical value for the coal’s stickiness, called a CSN index. The coolest temperature that initiated bonded deposits, Tai, was also recorded. The OEM then compiled a regression

Fig. 2.4 Parity plots for pairs of subgroupings of the database where one group is the regression database, and the other is a validation database and vice versa.

26

Process Chemistry of Coal Utilization

database on ten subbituminous and hv bituminous coals and their test results, including the proximate and ultimate analyses. Attempts to develop accurate regressions with combinations of standard coal properties were unsuccessful. They then developed an enthalpy balance for the test to estimate the coal thermal history and used FLASHCHAIN to predict the devolatilization behavior throughout the test. Eight predicted quantities were screened as predictors: the predicted yields of tar precursors (metaplast), tar, gas, CO2, H2O, CO, GHCs, and the number-average molecular weight of tar, Mn. In turn, each predictor was evaluated at 263°C, 279°C, 296°C, 313°C, 329°C, 346°C, 363°C, 379°C, 396°C, 410°C, and 430°C. Only the sum of the oxygenated gas yields and Mn survived the initial screening calculations because the remaining product yields were similar among coals that had different adhesion characteristics. The best single-variable regressions are collected in Table 2.3. In the correlations, the numbers in parentheses are the temperatures at which the yields and Mn were evaluated. For the CSN index, YOXY at different temperatures and Mn have comparable and remarkably good performance, whereas for Tai, YOXY at different temperatures and YTAR have comparable and very good performance. Although the predicted deposition potentials were never formally evaluated, these regressions were able to flag several coals with high adhesion potentials, without the complications of ongoing lab testing. Regarding why the oxygenated gas yields from devolatilization were such strong predictors, recall that adhesion is governed by coal softening and stickiness, which are aspects of the depolymerization of coal macromolecules. Coal must first disintegrate into fragments before it can soften and ultimately expel tarry devolatilization products. Once the fragments have formed, tars can only be swept away in a stream of noncondensable gases because the diffusion rates of such large molecules through microporous solids like coal are extremely slow. This latter requirement connects coal depolymerization to the production of oxygenated gases because the oxygenated gases are the first noncondensable devolatilization product to form. Oxygenated gases are the first indicator of extensive depolymerization, and also the carrier stream for the first tars to escape their parent particles.

Table 2.3 Single-variable regressions for two adhesion indices with predicted devolatilization behavior. Adhesion index

Correlation

r2

CSN

CSN ¼ 5.561  0.703YOxy(430) CSN ¼ 4.880  0.927YOxy(370) CSN ¼ 32.36  0.124Mn(430) CSN ¼ 4.956  1.219YOxy(329) CSN ¼ 4.581  2.077YOxy(279) Tai ¼ 219.8 + 16.14YTar(430) Tai ¼ 337.7 + 8.96YOxy(430) Tai ¼ 349.8 + 13.20YOxy(350) Tai ¼ 347.9 + 14.62YOxy(329) Tai ¼ 352.2 + 24.72YOxy(279)

0.928 0.927 0.915 0.942 0.926 0.868 0.890 0.873 0.850 0.863

Tai (°C)

Statistical prediction models

27

The next case study was compiled from a database of 21 tests in a 0.7 MWth pilotscale flame that was configured to recreate the emissions from commercial T-fired furnaces. The first goal is to predict the CQ impacts on NOX emissions from eight tests with eight different coals at the same staging level and excess O2 level and then to broaden the predictions for variations in staging level and excess O2 in 13 additional tests with three coals. The eight coals in the CQ suite covered ranks from subbituminous through medium volatile (mv) bituminous. The predictors are aspects of devolatilization behavior from FLASHCHAIN, with transformations that account for the instantaneous conversion of primary tar into soot and complete secondary volatiles pyrolysis. As primary volatiles are rapidly exposed to elevated temperatures at the inlets to coal flames, secondary pyrolysis products are the fuels that actually burn in the near-burner region. The predictors are the predicted yields of char, gas, and soot; the fractions of coal-N retained in char and soot, and two variables specified from the complete distributions of secondary products. As a scale for the calorific values of the volatiles from different coals, an adiabatic flame temperature, TAD, was evaluated for the thermochemical equilibrium of the primary air and volatiles streams, omitting char oxidation. As a rough scale for staging in the near-burner region, a stoichiometric ratio, SRPR, was evaluated from the primary air and whole coal streams. This parameter has the same value for the eight tests in the CQ test series but varies in the database for staging and excess O2. The only standard coal properties screened in the regressions were the C- and N-contents from the ultimate analyses. The proposed correlating function for NOX emissions resolves the CQ impacts from the impacts of the operating conditions as a product of separate functions, as follows: NOX ¼ ðNOX ÞBASE f ðO2 , S%, etc:Þ

(2.4)

where (NOX)BASE is an estimate for the emissions for the subject coal at baseline firing conditions, and f is a function that adjusts the emissions for deviations in the staging level and excess air from the baseline conditions. In this flame, the only staging variation was through an OFA adjustment. The basis for this functional form is shown in Fig. 2.5. The measured emissions across the range of excess O2 increase continuously for progressively greater O2 levels for all four coals, but the absolute magnitudes also increase for progressively higher rank in this staged combustor. Simply normalizing the measured values by the emissions for 3.55% O2 collapses the family of curves into a common quadratic or cubic polynomial, f(O2). A regression of the eight CQ tests will determine (NOX)BASE, and a separate regression of the 13 tests on operating conditions will determine the function f(S%, O2). For the CQ suite, predicted char yields varied from 39 to 51 daf wt.%; soot yields varied from 23 to 42 daf wt.%; char-N fractions varied from 0.49 to 0.57; and TAD varied from 1525°C to 1867°C, with uniform SRPR. The best single predictor was the partitioning of coal-N into volatile- and char-N, albeit with r2 below 0.30. The best coefficient for two-variable regressions (C, N) was 0.72, and for three-variables (N, NfCHAR, YSOOT) it increased to 0.84. The best four-variable regression had a coefficient of 0.89 and is based on N, NfCHAR, YTOT, and TAD. Higher-order regressions had

28

Process Chemistry of Coal Utilization

Fig. 2.5 (Left) Measured NOX emissions for four coals over a range of excess O2 levels; and (right) the emissions normalized by the value at 3.55% O2 for each coal.

Fig. 2.6 Parity plot for the regression of the CQ database.

greater r2-values but were avoided because there were only eight coals in the regression database. The performance of the four-variable regression of the CQ database is apparent in the parity plot in Fig. 2.6 and in the standard deviation (std. dev.) on the correlated NOX emissions of 16.5 ppmv, which is satisfactory for the range of measured emissions from 170 to 340 ppmv.

Statistical prediction models

29

To interpret the database on staging conditions, (NOX)BASE was evaluated from the regression of the CQ database for the three tested coals, and the predictors were S%, which varied from 0% to 25%; O2 which varied from 2.5% to 3.5%; and TAD, which varied from 1645°C to 1893°C. With a correlation coefficient of only 0.68 and an std. dev. of 46 ppmv across a range of measured values from 170 to 410 ppmv, this regression is not accurate to within useful quantitative tolerances. An alternative approach to rectify this situation will be presented in Chapter 3.

2.4

Neural nets

Whereas statistical regressions have been used for as long as people have compiled databases on coal processing, neural nets are a relatively recent alternative means to developing predictive models (Haykin, 2009). One can think of a neural net as a primitive computing system formed from some number of layers of neurons. Neurons are the processing units that take input values from a lower layer, transform them, and then pass them on to a higher layer. Fig. 2.7 illustrates a three-layer neural net with an input layer at the bottom, a hidden layer in the middle, and an output layer at the top.

Fig. 2.7 Three-layer neural net with an input layer at the bottom, a hidden layer in the middle, and an output layer at the top. Reproduced with permission from Murty BSN, Ravikumar YVL, Dutt, NVK. Estimation of coal conversion by supercritical solvents using the method of neural nets. Fuel 1997;76:165–68, Elsevier.

30

Process Chemistry of Coal Utilization

Layers between the bottom and top layers are denoted as “hidden.” Each neuron connects with all neurons in the layers above and below. The network also contains a bias neuron that passes some constant value to the neurons in the layers above it. In Fig. 2.7, the bias neuron is in the bottom layer. Each neuron contains some transfer function that attenuates the input value before it is passed onto the layer above. In addition, the output values from any neuron are weighted by multiplicative factors, denoted as Wij and wij in Fig. 2.7 where i is an index on a neuron, and j is an index on the receiving layer. These “weights” distinguish two operating modes for the network. In the learning mode, output values from a database are used to fine-tune the weights to match the input variables. In this mode, information flows downward from top to bottom to “train” the neurons to mimic a certain database. In this so-called “back propagation” mode, errors evaluated from each neuron are minimized by adjusting the weights needed to match the input data. This analysis is repeated for numerous input cases until the average error on each input variable over all cases satisfies a specified tolerance. In the prediction mode, the weights are unchanged while information flows from the input values, upward through the layers to predict the ultimate output value(s). Compared to statistical regressions, neural nets have the advantages of being able to process a multitude of input variables, to represent highly nonlinear relations among input and output variables, and to better tolerate noisy and uncertain input data. However, their database requirements are very large, and field testing usually requires specialized personnel, which is expensive. These concerns are receding rapidly as more utility operators install distributed control systems (DCS) that continuously record furnace operating conditions and performance parameters. Provided that a furnace is equipped with a DCS that compiles data on effective predictor variables and the output variables, the data requirements for neural nets are manageable. These concepts are illustrated further in the first case study, which develops a network to predict coal extract yields from a diverse assortment of coals in a wide variety of solvents over broad ranges of temperature and pressure (Murty et al., 1997). This analysis used the three-layer network in Fig. 2.7, and a sigmoid transfer function, [1 +e x]1 at each neuron. The input variables were characteristics of the solvents (critical temperatures and pressures, TC and PC); test conditions (TE, PE), and coal properties (H/C and VM). The only output variable is the fractional coal extract yield (Y), which varied from 0.19 to 0.80 in the database. The database was compiled from reported extract yields for 29 coals in 15 solvents at temperatures from 375°C to 550°C and pressures from 10 to 30 MPa. Due to the simplicity of the network in this particular case study, the ultimate prediction tool could be expressed explicitly as follows: 2

9 > > > > =

8 > > > >
> X > > 5 4 j¼1 0 > > > > wij xi ; :1 + exp  i¼1

(2.5)

Statistical prediction models

31

where xi0 are the input variables expressed as a difference with the mean for the database normalized by the std. dev. (Murty et al., 1997); Wji are weighting factors among hidden and output nodes; and wij are weighting factors among input and hidden nodes. The correlation coefficient is 0.987, and the average absolute deviation in the predicted yields is 3.5%. It is interesting that the assigned weights in the prediction model indicate that predictors representing operating conditions, solvent properties, and the CQ impacts made comparable contributions to the predicted extract yields. This is certainly not a reaction system whose yields are easy to predict, especially in 15 different solvents. The next case study revisits the goal of predicting the extent of coal conversion in a full-scale furnace with a neural network, rather than a multivariable regression. It used a 300 MW T-fired furnace equipped with a low-NOX concentric firing system implemented over five fuel injection elevations and one OFA port in each corner (Hao et al., 2003). The database was compiled from 21 tests in which the distribution of secondary air, OFA levels, and coal quality were varied. The goal is to predict LOI, which is the weight percentage of UBC in ESP fly ash, for any coal type at any furnace operating conditions. The “tests” were periods in which the furnace operation was stable, and all relevant operating conditions could be recorded. The various groups of tests were directed at secondary air distribution patterns, OFA variations, burner tilt levels, excess O2, pulverizers-in-service, and furnace load. They were characterized with 21 predictors: load; 5 secondary air damper settings; 5 coal feedrates; primary air feedrate; injector nozzle tilt, OFA damper position, fuel-air damper position, pressure drop from the wind box to the furnace, and excess O2. Coal quality was characterized with three predictors from the proximate analyses and the calorific values. The LOI predictor was developed from a three-layer neural net with 21 input neurons, 24 neurons in one hidden layer, and a single output neuron for LOI. Twenty tests were used to train the network to a mean square error of less than 105. Measured LOI levels varied from 1.5 to 8.2 wt.%. The agreement between predicted and measured LOI levels was exact in 17 of the 20 learning tests, and none of the discrepancies among the other three cases exceeded 0.1 wt.%. One additional test was analyzed in the prediction mode, for which the predicted and measured LOI values were nearly the same at 2.42 vs 2.35 wt.%. Obviously, many more validation cases would need to be examined before the accuracy of the network was firmly established. Nonetheless, this case clearly demonstrates that neural nets can handle a multitude of input variables. This capability enables them to simultaneously track multiple influences on the output variable, regardless of their relative magnitudes. LOI emissions are not determined by a single dominant factor with attenuation from an assortment of minor influences. Rather, it is the cumulative result of numerous disparate factors, some acting independently and some in concert, while the fuel streams burn as they move through different furnace regions. In this case study, the network simply utilized all the in-furnace controls that are accessible to an operator plus only the most standard coal properties, to “learn” how LOI responds to each conceivable furnace adjustment. In contrast, a multivariable regression would saturate to some stable level of performance with only a handful

32

Process Chemistry of Coal Utilization

of the potential predictors, without any means for refinements that could capture second-and third-order impacts. In general, the resolution of a very large number of factors is the main advantage of neural nets in predicting the performance of coal-fired furnaces. The next case study was conducted by the same research institute as the previous LOI prediction analysis (Zhou et al., 2004). It used a 600 MW T-fired furnace equipped with a low-NOX concentric firing system implemented over six fuel injection elevations, with five secondary air ports, and two OFA ports in each corner. The database was compiled from 12 tests in which the distribution of secondary air, burner tilt, OFA levels, load, and coal quality were varied (Zhou et al., 2004). Two dissimilar hv bituminous coals were represented, one of which had 60% more N than the other. The goal is to predict average NOX emissions over several hours of operation across the full domain of furnace conditions and CQ. The “tests” were periods in which the furnace operation was stable, and all relevant operating conditions could be recorded off a DCS. The various groups of tests were directed at secondary air distribution patterns, OFA variations, burner tilt levels, furnace load, and CQ. They were characterized with 23 predictors: load; total coal feedrate; 5 coal distribution settings; 5 primary air feedrates; total air flowrate; injector nozzle tilt; 5 secondary air damper settings; 2 OFA damper positions; pressure drop from the wind box to the furnace, and excess O2. Coal quality was characterized with partial proximate and ultimate analyses and calorific values. The NOX predictor was developed from a three-layer neural net with the 23 input variables on operating conditions listed previously, plus C, H, N, O, VMar, and HHV for CQ. Consequently, the neural network contained 29 input neurons, 31 neurons in one hidden layer, and a single output neuron for NOX emissions. Eleven tests were used to train the network to a mean square error of less than 105. In the database, peak furnace temperatures varied from 1380°C to 1500°C, and measured NOX emissions varied from 620 to 907 mg/Nm3@ 6% dry O2. As expected, NOX emissions decreased for progressively lower loads. Whereas the agreement between calculated and measured NOX emissions for the training tests was virtually exact, a single validation case that used the neural net in its prediction mode gave 681 mg/Nm3@ 6% dry O2 vs a measured value of 659, which gives a relative error of 3.3%. A companion analysis of LOI emissions from this furnace gave 1.22 wt.% vs a measured value of 1.16 wt.% (Zhou et al., 2004). Unfortunately, no single validation case can unequivocally portray the accuracy of this neural network. As the number of predictors increases, computational burdens can become excessive in the implementation of neural nets. These burdens can be managed by discarding inconsequential variables from the networks. In the development of a network to predict NOX emissions from a commercial CFBC, Liukkonen et al. (2011) analyzed 10,000 process records of 42 operating conditions from a DCS that covered 104 days of continuous operation. They used self-organizing maps to identify five distinct operating modes: low steam flow/low bed temperature; low steam/unstable temperature; medium steam/medium temperature; medium steam/medium temperature with unstable fuel and air flows; and high steam/high temperature. For each operating mode, this group identified no more than 10 predictors that met their

Statistical prediction models

33

performance criteria. Although the 10 best predictors were not the same for each mode, adding additional predictors did not improve the performance in any mode. And all the computational burdens were manageable. Unfortunately, the group apparently had no access to standard coal properties or any CQ predictors, so their networks functioned solely on the basis of the furnace operating conditions.

2.5

Summary

With the widespread adoption of continuous monitoring via DCS at coal-fired furnaces, there is little doubt that neural nets are fast becoming the prediction tool of choice to manage and optimize furnace operations. Indeed, only DCS can compile the enormous databases needed to train neural nets, as dedicated field testing is too specialized, too expensive, and too slow to meet these demands. Given a database of sufficient size, neural nets also have the capabilities to simultaneously process all the variables recorded by a DCS that pertain to an output variable. There is no need to understand in advance the connections among, for example, furnace load and excess O2 and NOX emissions. Once properly trained, a network will display this tendency and all other important tendencies among the input and output variables. Nor is it necessary or advantageous to use the most powerful predictor variables because using all available inputs is nearly impossible to improve upon. Neural nets also have the potential to identify connections of second- and third-order importance that multivariable regressions and other statistical methods would surely miss. Case studies for neural nets in applications with coal-fired furnaces certainly have not yet established the accuracy of the predictions, because they are limited to very few cases in which the network operated as a legitimate prediction model. But if the accuracy actually improves with progressively more training data, then it seems inevitable that nets will ultimately make predictions as accurately as they need to be to manage commercial furnaces. Of course, computational burdens can become intolerable for very large databases, but this concern has already been addressed with methods to screen for the minimum numbers of predictors that meet a specified performance criterion. Of course, there is a multitude of important applications in coal utilization technologies where the input data requirements of neural nets are prohibitive. These situations often involve coals behaving as “bad actors,” whereby problems arise only with a small percentage of the coals processed in the same ways. In this chapter, they were illustrated with the case studies on autoignition and burner deposits. Now that more countries participate in the world coal trade, especially developing Asian countries, such situations will only become more common. Multivariable regressions are well-suited to these situations because they usually identify the CQ impacts given data on only a few dozen different coals. Their performance can be enhanced by using more powerful predictors, such as those with direct connections to the coal constitution and to coals’ devolatilization behavior. In these situations, analysts with sharp intuitions on coal conversion chemistry hold a marked advantage.

34

Process Chemistry of Coal Utilization

References Cloke M, Lester E, Gibb W. Characterization of coal with respect to carbon burnout in p.f.-fired boilers. Fuel 1997;76:1257–67. Hao Z, Qian X, Cen K, Fan J. Optimizing pulverized coal combustion performance based on ANN and GA. Fuel Process Technol 2003;85:113–24. Haykin S. Neural networks and learning machines. 3rd ed. Upper Saddle River, NJ: Pearson Education; 2009. Liukkonen M, Heikkinen M, Hiltunen T, Halikka E, Kuivalainen R, Hiltunen Y. Artificial neural networks for analysis of process states in fluidized bed combustion. Energy 2011;36:339–47. Murty BSN, Ravikumar YVL, Dutt NVK. Estimation of coal conversion by supercritical solvents using the method of neural nets. Fuel 1997;76:165–8. Niksa S. FLASHCHAIN® theory for rapid coal devolatilization kinetics. 3. Modeling the behavior of various coals. Energy Fuel 1991;5:673–83. Niksa S. FLASHCHAIN® theory for rapid coal devolatilization kinetics. 4. Predicting ultimate yields from ultimate analyses alone. Energy Fuel 1994;8:659–70. Niksa S. Process chemistry of coal utilization: impacts of coal quality and operating conditions. London, UK: Woodhead Publishing, Elsevier; 2020, ISBN:978-0-12-818713-5. Sathyanathan VT, Mohammad KP. Prediction of unburnt carbon in tangentially fired boiler using Indian coals. Fuel 2004;83:2217–27. Tomei GL, editor. Steam its generation and use. 42nd ed. Charlotte, NC: The Babcock and Wilcox; 2015. Zhou H, Cen K, Fan J. Modeling and optimization of the NOX emission characteristics of a tangentially fired boiler with artificial neural networks. Energy 2004;29:167–83.

Heuristic prediction schemes

3

Nomenclature CQ L N fCHAR O2 (NOX)BASE (NOX)MAX (NOX)REF ΔNOX S% TAD W∞

coal quality furnace load, % fraction of coal-N in char excess O2 at a furnace exit measured NOX emissions for a baseline furnace condition hypothetical maximum NOX emissions from screening or baseline fuel at baseline operating conditions estimated NOX emissions for a baseline fuel at conditions of the internal NOX database difference between estimated NOX emissions for a screening fuel and measured NOX for a baseline fuel staging level as a percentage of total furnace air adiabatic temperature for a volatiles/primary air flame, °C total volatiles yield, daf wt.%

Greek symbols αNOx υO 2

CQ predictor for NOX emissions from a screening or baseline fuel air requirement for fuel combustion in a burner belt

Subscripts BASE REF

pertaining to the fuel or conditions in a baseline furnace dataset pertaining to conditions in the internal NOX database

This chapter covers heuristic prediction schemes, which try to make statistical regressions “smarter” by factoring in at least some of the most important relations that determine a target output variable. These relations are almost always at the phenomenological level, rather than the mechanistic. For example, NOX emissions diminish for progressively deeper extents of staging. This relation could be incorporated into a prediction model as an explicit mathematical factor on NOX that was inversely proportional to the extent of staging, S%. Or, it could arise as some functional dependence on the stoichiometric ratio for the near-burner region, whose value diminishes for greater S%. Or, it could even be implemented in the form of a much simpler combustor flowfield than the one in the actual furnace under consideration. From the perspective of the process chemistry, a 1D plug flow system fed with coal and multiple air streams exhibits all the essential tendencies on NOX emissions of a large furnace, but without Process Chemistry of Coal Utilization. https://doi.org/10.1016/B978-0-323-89959-8.00002-1 Copyright © 2022 Elsevier Ltd. All rights reserved.

36

Process Chemistry of Coal Utilization

the complications of large spatial and thermal gradients. Any scheme that highlights the process chemistry has the potential to depict the CQ impacts. Given the right supporting information and a robust calibration protocol, even the simplest analogs for furnace flowfields can perform better as prediction models than statistical regressions.

3.1

NOXLOI Predictor

The bulk of this chapter is based on a case study of a single software product called the NOXLOI Predictor because its performance in emissions predictions has been fully validated and very well documented (Niksa et al., 1997, 1999, 2003a; Sun et al., 2003). This package was developed in the late 1990s by the Electric Power Research Institute (EPRI) as a tool for fuel procurement, not to manage furnace emissions by regulating the furnace firing conditions. It was distributed to over 80 American utilities and a few international firms and has been useful to those in the utility industry concerned with fuel supplies, spot purchases, NOX compliance strategies, environmental engineering, day-to-day operations, and fly ash utilization. The NOXLOI Predictor predicts NOX, UBC, and LOI emissions for a broad range of fuel quality, furnace firing configurations, and operating conditions. Unburned carbon is either expressed directly as a percentage of the combustibles into the furnace or as the weight percentage of combustibles in fly ash, which is called loss on ignition or LOI. This software focuses on the CQ impacts. It does not try to predict the absolute NOX emissions, LOI, and UBC by simulating the various mechanisms within a specific boiler. Rather, the prediction is based on measured NOX data and LOI or UBC data for a baseline coal or coal blend fired under known conditions. Given this data, the program calculates the expected change in emissions among the baseline coal and one or more different fuels when they are fired under identical conditions. Estimates from this software should be recognized as first estimates that could be refined in subsequent calculations that account for the effects of different loads, excess O2 levels, and other firing adjustments after a fuel switch. The overall structure of the calculation is simple: the user provides a few measured values of NOX and LOI emissions from a particular boiler and describes the major furnace specifications, including furnace rating, firing configuration, type of NOX control technology, grind size, and an index on the furnace capacity. The user also enters the proximate and ultimate analyses for the coal that was fired while the baseline data were recorded, as well as for any coals and multifuel blends, including coals blended with petroleum cokes (petcokes) or biomass, that will be screened in the calculations. The computer program then predicts the NOX and LOI emissions for the set of fuels being screened when they are burned under the same firing conditions that were used when the baseline data were collected. This information enables fuel procurement specialists to refine their cost estimates for a fuel switch to account for variations in emissions and the associated costs to respond. Potentially troublesome fuels can be avoided altogether. Boiler operations engineers use the emissions forecasts to anticipate boiler adjustments and changes in the firing conditions needed to bring a new fuel online in the shortest possible time. Environmental specialists use the information to assess the cumulative impact of fuel

Heuristic prediction schemes

37

switches across a regional utility operation. Ash utilization managers may be able to anticipate changes in the salability of their ash and regulate their landfill requirements and sales accordingly.

3.2

Predicting CQ impacts on NOX emissions

Perhaps the most important aspect of the calculation sequence to predict how NOX emissions change with a fuel switch is how the CQ impacts are resolved from all the operational factors, many of which are stronger influences than CQ, per se. Numerous factors encompassing furnace design, burner operations, and fuel quality affect NOX emissions, especially the following: l

l

l

l

l

l

l

l

l

Burner firing configuration Burner design, including high turbulence or internally staged burners Burner settings, such as tilts and swirl vane settings Fuel composition Fuel/air distribution or maldistribution Air staging configuration, including CCOFA or SOFA Excess O2 level Furnace volume and residence time Furnace temperature profile

The NOX calculations focus on the CQ impacts, albeit only at face value because NOX emissions could not possibly be predicted within useful tolerances if the various boiler-related and burner-related factors were ignored. The predictions are based on measured NOX emissions for baseline conditions, and these measurements implicitly encompass the coal type, boiler characteristics, and operating conditions. These data are the means of incorporating all the boiler-related and burner-related factors into the prediction algorithms. Indeed if, for example, burner settings are varied or if the overfire air level is adjusted, then the baseline data entered into the calculations must cover the ranges of these variations. Note the crucial difference with the statistical regression proposed for NOX emissions from a pilot-scale flame in Chapter 2. In the former regression, the CQ impacts were implemented as a multiplicative factor, (NOX)BASE, according to NOX ¼ ðNOX ÞBASE f ðO2 , S%, etc:Þ

(3.1)

where (NOX)BASE is a prediction model that estimates emissions for any coal at baseline firing conditions, and f is a function that adjusts the emissions for deviations in the staging level and excess air from the baseline conditions. Whereas the predicted CQ impacts from the regression of the CQ database were reasonably accurate, the regression for f was not. In the NOXLOI Predictor, the fractional change in NOX emissions due to a fuel switch is predicted with the following functional relation:   ðNOX ÞBASE NOX  ðNOX ÞBASE ¼ ΔNOX ¼ gðCQ variablesÞf ðNOX ÞREF

(3.2)

38

Process Chemistry of Coal Utilization

where (NOX)BASE is a measured emissions value for specified baseline operating conditions and a baseline fuel sample; g is a prediction model that uses a selection of CQ variables for both the candidate replacement fuel and the baseline fuel sample; and f is a function that rescales the NOX estimate from reference database conditions to operating conditions in the subject boiler. Given the baseline emissions data, the prediction model calculates the expected change in emissions when changing from the baseline coal or coal blend to another fuel when it is fired under identical conditions. The other fuels in this calculation are called screening fuels. For example, if the baseline dataset includes NOX at full load with several different O2 levels, then NOX emissions from the screening fuels will be predicted for each of the O2 levels. Similarly, if the baseline dataset depicts NOX as a function of load, then the predictions for the screening fuels will also depict the load dependence. The one-to-one correspondence is the same for burner tilts and any other variable furnace condition, provided that all other conditions are uniform across each set of baseline data. However, there is an important caveat. The calculations do not consider whether all the screening fuels could actually be fired at the baseline firing conditions. For example, ash-related effects with certain fuels may entail changes in the burner tilts in a Tfired furnace, and varying tilts may affect NOX. The calculations neglect such considerations, and therefore further investigation and engineering judgment are often needed to refine the raw NOX predictions. Nevertheless, the predictions remain useful as relative indications of how fuel switching will affect NOX emissions and are the starting points for more refined estimates that account for the operating characteristics of particular furnaces. The NOX prediction model incorporates two components: (1) FLASHCHAIN to estimate the thermal decomposition of coal under flame conditions; and (2) a database of pilot- and full-scale NOX measurements for various firing configurations and operating conditions. FLASHCHAIN is used to predict two critical characteristics that relate fuel properties to NOX emissions. First, it predicts the total volatiles yield from the fuel while it is being heated under flame conditions, where heating rates can approach 100,000°C/s and temperatures approach 1650°C. The total weight loss under flame conditions typically exceeds the proximate volatile matter contents by 20% to 100%, which explains why NOx emissions do not correlate very well with fuel ratios determined from the ASTM proximate analyses. The second prediction from FLASHCHAIN is the partitioning of fuel nitrogen among gaseous pyrolysis products and char. This partitioning is important because the gaseous nitrogen compounds are the ones that can be managed by aerodynamic NOX abatement strategies that regulate mixing to achieve relatively reducing conditions at moderate temperatures in burner belts. In addition, the distribution of volatiles as molecular species from FLASHCHAIN is used to estimate a nominal flame temperature in the burner belt region, TAD. The flame temperatures are used to refine the NOX predictions for selected firing configurations, and also to assign the pretreatment conditions in the char oxidation submodel in the LOI prediction algorithms. The emissions from coal blends or multifuel blends are not assigned from an average set of pseudo-fuel properties. Rather, they are developed from the predicted volatiles yields and compositions and the nitrogen partitioning from each blend

Heuristic prediction schemes

39

component. Appropriate weighting functions for yield, product distribution, nitrogen release, and calorific value are used to assign the devolatilization behavior of the blend that, in turn, is the basis for the predicted NOX emissions. The second major element of the predictions is a database of NOX emissions from pilot- and full-scale furnaces with as many different fuel samples as could be found. First, datasets were collected for the NOX emissions for different coals, petcokes, and biomass-fired under the same conditions in the same boiler or pilot-scale flames, for as many different firing configurations as available. This database was then subcategorized according to the firing configuration (wall- or T-fired), furnace volume, and NOX control technology (low-NOX burners, SOFA, CCOFA). Then, regressions of NOX data in the subcategories were developed in terms of the FLASHCHAIN results. These regressions were then algebraically inverted to estimate reference NOX emissions for any baseline fuel given only the FLASHCHAIN results and enough information to identify a particular subcategory of firing configurations. The CQ impacts in the database correlated with FLASHCHAIN predictors are related to a subject full-scale boiler with another calibration procedure. The measured baseline NOX relates the database correlations to a specific full-scale unit, using a reference NOX level that the baseline coal would produce under the database conditions. In essence, the baseline NOX level is used to calibrate the correlation for the internal NOX database. The internal databases for T-fired and wall-fired furnaces are sufficient to make reliable predictions. However, the data on cyclone-, top-, arch-, or turbo-fired furnaces were insufficient for this compilation. The calculations to relate the predicted devolatilization behavior and TAD to the change in NOX emissions for the screening coals incorporate several established tendencies in the ways that CQ affects NOX emissions. Two are that NOX emissions increase for progressively greater coal-N levels and that aerodynamic NOX abatement becomes more effective under lower O2 concentrations. Both tendencies are implemented as a hypothetical maximum NOX concentration from the burner belt region, (NOX)MAX, according to ðNOX ÞMAX ¼

c1 N υO2

(3.3)

where υO2 is the stoichiometric O2 requirement to burn out coal, based on its ultimate analysis; and c1 is a constant. The greater the O2 requirement for coal combustion, the lower the SR will be in the burner belt, and NOX production is inhibited under fuelrich conditions. The maximum NOX level is directly proportional to coal-N. As only volatile-N is amenable to aerodynamic NOX abatement, the NOX emissions from each fuel are resolved into independent contributions from volatile-N and char-N, according to αNOX ¼

c2 + c3 N fCHAR ðNOX ÞMAX W∞

(3.4)

where W∞ is the ultimate yield of volatiles, and NfCHAR is the char-N fraction from FLASHCHAIN. Eq. (3.4) ensures that char-N is converted into NOX at maximum

40

Process Chemistry of Coal Utilization

efficiency, whereas the contribution for volatile-N is inversely proportional to the total volatiles yield. Values of αNOx for the baseline and screening fuels express their NOX production potential. Moreover, the αNOx-value for a baseline fuel is used as the regression variable in a single-variable prediction model for the database of NOX emissions for standardized combustion conditions. This estimate for NOX under the conditions for the database is denoted by (NOX)REF. The ultimate change to the measured, baseline NOX emissions for a screening fuel is given by    ðNOX ÞBASE C5  ΔNOX ¼ c4 αNOX  ðαNOX ÞBASE ðNOX ÞREF

(3.5)

where ΔNOX is the difference between the predicted NOX emissions for the screening fuel and the measured emissions for the baseline fuel, and (αNOx)BASE is the value from Eq. (3.4) for baseline fuel properties. Hence, the change in emissions from a fuel switch accounts for the different NOX production potentials of the screening and baseline fuels, through the difference in the αNOx-values in square brackets; and also for the difference in the NOX production potential of the subject furnace compared to the reference conditions in the NOX database, through the ratio (NOX)BASE/(NOX)REF in parenthesis raised to the power c5. Eqs. (3.3)–(3.5) are implemented as written for T-fired furnaces. For wall-fired furnaces, TAD is also incorporated. The same definition for (NOX)MAX in Eq. (3.3) is retained, but αNOx is expanded as follows:   c6 c9 N αNOX ¼ + c7 fCHAR ðNOX ÞMAX + c8 exp  ð1N fCHAR ÞðNOX ÞMAX TAD

(3.6)

where the first term on the right for volatile-N conversion explicitly factors in NfVOL (as 1 NfCHAR), and a third term accounts for the contribution from thermal NO; i.e., conversion of N2 that enters the flame as air. The definition of ΔNOX in Eq. (3.5) also pertains to wall-fired furnaces, albeit with different numerical coefficients. To validate the NOX predictions from the NOXLOI Predictor, measured stack NOX emissions are needed for the baseline and screening coals for the same firing conditions. Depending on the boiler type, the firing conditions would include load, pulverizers in-service, staging level, burner tilts, burner damper settings, and excess O2. If test data for different fuels are not available under identical firing conditions, then the results are adjusted to put the comparison on a common basis, to the extent possible. For instance, a difference in the flue gas O2 concentration of 1% can change NOX on T- and wall-fired boilers by 30–50 and 60–90 ppm, respectively. Likewise, a change in burner tilt can impact NOX on the order of 1 to 2 ppm per degree of tilt change. In some cases, small changes in operating parameters may exert an impact that is equal to or greater than the fuel quality impacts. Over several years EPRI was able to enlist the support of its utility members to provide tens of datasets on measured NOX emissions from full-scale furnaces to validate predictions from the NOXLOI Predictor. A parity plot of a large selection of the

Heuristic prediction schemes

41

Fig. 3.1 Parity plot on NOX emissions from commercial (□) T- and (▪) wall-fired utility furnaces from the NOXLOI Predictor. None of the measured values were in the database used to evaluate coefficients in regressions in the prediction model (Niksa et al., 1999).

predicted versus observed NOX emissions for a wide range of coal rank for T- and wall-fired furnaces appears in Fig. 3.1. None of these cases were in the original database used to determine the regressions in the first place. In general, the predicted values are within 10 to 20 ppm of the measured values for coals whose ranks represent subbituminous, hv bituminous, and mv bituminous, which would be adequate for most applications. The differences between the baseline emissions and the predicted emissions for multiple screening coals across ranges of load appear in Fig. 3.2 for two commercial furnaces. Emissions were available from each unit for a baseline coal and two screening coals as a function of furnace load. Six emissions measurements for the baseline coals characterized NOX as a function of load over the top one-third and one-fifth of the furnace load range for the furnaces. All these data for the two baseline coals were used to predict NOX as a function of load for both pairs of screening coals over the respective load ranges. There is a one-to-one correspondence between the operating conditions represented by the baseline dataset and the operating conditions that pertain to the NOX predictions for the screening coals. In this case study, furnace load is the primary operating variable while the other operating conditions are assumed to be the same among the tests with the baseline coal and both screening coals. The predictions are evaluated with test data in Fig. 3.2 and compared to the baseline dataset that was input into the NOX prediction model. The smooth solid curves are spline fits to the predicted values for the operating conditions of each baseline data point and the screening coal properties.

42

Process Chemistry of Coal Utilization

0.7 143 MW T-Fired

Baseline Emissions

NOX, lb/MBtu

0.6

Coal E Coal F

0.5

Coal D

0.4

Coal C Baseline Emissions

0.3

0.70

0.75

194 MW Wall-Fired

0.80 0.85 0.90 Fraction of Full Load

0.95

1.00

Fig. 3.2 Evaluation of predicted NOX emissions from commercial (upper grouping) T- and (lower grouping) wall-fired furnaces fired with two screening coals each across a furnace load range (Niksa et al., 2003a). Baseline NOX data for hv bituminous coal appear as dashed curves with ▀.

In the wall-fired unit, the baseline NOX emissions increased from 0.30 to 0.32 lb/MBtu as the load was increased from 80% to 100% of the furnace rating. Over the same load range, the NOX emissions from screening coal D were predicted to increase from 0.41 to 0.44 lb/MBtu, versus measured values from 0.41 to 0.44 lb/MBtu. The NOX prediction model accurately predicts the 37% increase in NOX over this load range due to the switch to coal D. For coal C, NOX emissions were predicted to increase from 0.36 to 0.38 lb/MBtu, versus measured values from 0.36 to 0.40 lb/MBtu, and the 18% to 20% increase in NOX compares well to the measured increase of 20% to 25% over this load range. The baseline NOX emissions for the T-fired unit exhibit the opposite tendency to decrease for progressively greater loads. As this trend is expressed by the baseline dataset, the predicted NOX emissions for both screening coals in this unit also decrease for greater load levels. The prediction model does not rely on reaction mechanisms to reproduce the trend in a baseline dataset; rather, any distinctive feature in the baseline data will propagate through the internal correlations into the predictions for the screening fuels. Sensitivities to the variations in the conditions are not necessarily the same for baseline and screening fuels, but the primary tendency will be apparent with the screening fuels. The predicted emissions from Coal F are 10% to 11% lower than the baseline emissions, which is within the measurement uncertainty over the full load range in the tests. The predicted emissions for Coal E are essentially the same as the baseline emissions, which is consistent with the data except at 90% to 96% load. The case studies in Fig. 3.3 evaluate predicted NOX for a coal blend in a 500 MW T-fired boiler fitted with a low-NOX concentric firing system, and for a beneficiated

Heuristic prediction schemes

43

Fig. 3.3 Evaluation of predicted NOX emissions from (left) a commercial furnace fired with a coal blend and (right) a pilot-scale flame fired with a beneficiated coal (Niksa et al., 1999). Baseline NOX data appear as dashed curves with ▀.

coal in a 0.6 MWth pilot-scale flame. For the T-fired furnace, the baseline data were first fit with the dashed regression line that could be evaluated at 6% and 7% O2. Then, the NOX prediction model was run at these two O2 levels to predict NOX versus exit O2 for the 25:75 screening coal blend. If the line through both of the predicted NOX emissions were extended across the full O2 range, it would lie very close to a regression of all the available data for the blend, so the predictions are reliable. The second case in Fig. 3.3 uses a beneficiated coal whose properties resemble subbituminous coals’ in both high volatility and high coal-O. Whereas the baseline emissions from a hv bituminous coal increase from 210 to 300 ppm across the O2 range, the predicted emissions for the beneficiated coal are much lower, increasing from only 130 to 190 ppm over the same O2 range. Most of the difference is because the beneficiated coal contained only about half the coal-N as the baseline coal. The predicted emissions are also less sensitive to O2 variations than the baseline coal, as evident in the data. The performance with opportunity fuels is highlighted in the evaluation in Fig. 3.4 for blends of coal with pet coke and biomass. For the T-fired boiler cofiring petcoke, the baseline coal is a subbituminous and the unit cofires petcoke at up to 20% of the heat input. The predictions are in good agreement with the measured NOX levels for the petcoke blend. At low exit O2 levels, the predicted NOX levels are almost exact, whereas, at the highest O2 levels, the predictions are somewhat higher than measured but still reasonable. For the wall-fired unit firing blends of biomass on coal, the biomass was a band sawdust and circular sawdust, and the tests monitored as much as half the feed as biomass. The NOX predictions are based on the baseline emissions of 575 ppm for the 20:80 biomass:coal blend. They are accurate across the entire range of blend compositions, albeit within the considerable scatter in the measured emissions for moderate biomass loadings.

44

Process Chemistry of Coal Utilization

Fig. 3.4 Evaluation of (□) predicted NOX emissions from commercial furnaces fired with (left panel) a (●) coal/petcoke blend and (right panel) a (●) biomass/coal blend. For the petcoke blend, baseline emissions appear as ▀ and dashed line. For the biomass blend, the 20% biomass baseline blend was the baseline fuel (Niksa et al., 1999).

Clearly, the prediction model for NOX emissions is accurate for essentially the complete range of CQ in commercial furnaces. The predictions are as accurate for blends of coal with other coals, petcokes, and biomass as for neat coals. Several important generalizations on this prediction model are presented after the section on LOI predictions.

3.3

Predicting CQ impacts on LOI emissions

The list of operational factors that affect LOI emissions includes those for NOX emissions, plus the fuel hardness, fineness, and top size; pulverizer performance; char density and swelling behavior; the intrinsic char oxidation reactivity; and a char’s propensity for thermal annealing and ash agglomeration. The multitude of factors affecting coal burnout makes predictions for LOI very difficult to make. There has been an intensive search worldwide for the causes of LOI during the last three decades, ranging from R&D work on control strategies from bench-scale testing to computer simulations of full-scale furnaces. This body of research is most remarkable for the sheer number of factors that have been found to affect LOI levels. Grind sizes, ash levels, and the variations among particle morphologies and molecular constitution are the most important fuel properties, both among coals and within a particular coal sample. Firing configuration, mixing intensity, flame temperature, oxygen level, fuel/ air maldistribution into burner arrays, and residence time are also important because they affect char burnout. Finally, the intrinsic oxidation rates of chars are now known to significantly diminish during the later stages of burnout, due to an interplay among several independent mechanisms (Niksa et al., 2003b). Whereas all these factors have been shown to affect LOI, their relative impact is difficult to rank-order, although it is already clear that no single effect determines the LOI level in a full-scale furnace. The NOXLOI Predictor estimates UBC from a mechanism that incorporates a variety of important fuel- and boiler-related factors, including coal rank parameters (such

Heuristic prediction schemes

45

as inertinite and ash contents and grinding behavior), operating conditions (such as fuel/air mixing rates), and char properties (such as yield, density, and reactivity). UBC is determined by modeling the combustion of individual fuel particles of many different sizes passing through a simplified furnace environment, then averaging the conversions over the particle size distribution (PSD) to assign one UBC level for the whole flame. The modeling comprises submodels of the grinding process, the furnace environment, and the combustion and burnout behavior of individual particles. As such, it illustrates prediction models based on radically simplified model analogs for an actual furnace environment. The LOI prediction model applies the strategy that LOI can be estimated by first calculating UBC, and then by converting this value to LOI with a coal’s ash content, assuming that none of the mineral matter in coal vaporizes during fly ash formation. More sophisticated approaches based on uneven partitioning of ash and/or UBC between bottom ash and fly ash were not incorporated but would be warranted for cyclone firing, where most mineral matter ends up in bottom ash. Notwithstanding these simplifications, the assignment of LOI values from UBC values is inherently uncertain because of the way LOI is measured. According to ASTM protocol, LOI is measured by burning away all combustible material from a fly ash sample and recording the change in weight as LOI. In actuality, inorganic material will also be vaporized in the test, particularly the sulfate forms. Such contributions to weight loss are the reasons that LOI determinations overestimate the actual combustible material. More accurate determinations based, for example, on elemental analysis with direct measurement of the carbon levels, show that the overestimation is usually no more than 0.5 wt.% but can be as much as 2 wt.%. The LOI values assigned from the UBC prediction model do not account for these ambiguities.

3.3.1 Fuel grinding submodels As different coals have different grinding behavior, it is impossible to predict the CQ impacts on LOI emissions without accounting for the changes in the PSD from the pulverizers when coals are switched. The LOI prediction model supports two options for pulverizer performance for all nonbiomass components, as follows: (1) Leave the grind size distribution unchanged from the baseline, assuming that the pulverizers can be adjusted to achieve the same grind with any of the screening coals or petcokes; or (2) Estimate how the size distribution of each screening coal or coke would differ from the baseline size fractions if the pulverizer settings stayed the same, assuming that variations among the Hardgrove Grindability Index (HGI) and calorific values of all coals or cokes indicate the true tendencies.

This choice must be considered carefully because the LOI predictions under these options can differ by up to a factor of two. In our evaluations, LOI predictions based on the estimated size distribution have usually been more accurate than those for the baseline grind fractions. If new grind fractions are selected, the software first calculates a change in pulverizer capacity assuming constant heat input into the boiler to maintain the furnace

46

Process Chemistry of Coal Utilization

rating. This calculation uses the as-received calorific value, which is estimated from the Dulong-Petit equation for coals and cokes, and from a separate correlation for biomass that rectifies the underestimation of the calorific values with the Dulong-Petit equation. The change in pulverizer capacity together with the differences in the respective HGI values of the baseline and screening coals and cokes determines a new fineness (assigned as wt.% under 200 US Standard Mesh) based on performance correlations of full-scale pulverizers. The estimated fineness determines a RosinRammler distribution for the screening coal or coke, assuming that the top-size fraction is unaffected by the fuel switch. This latter assumption—called the constant top-size model—envisions an ideal classifier that limits the size of the largest particles leaving the pulverizer, as would be achieved in practice by recycling more material to the grinding zone. This approach predicts small changes in the >50 mesh fraction even when fineness (10% a0 a1 a2 a3 a4 a5

Support for CFD and process simulation applications 2100

1.0

89

100

lignite

2000 0.8

Temperature, °C

lignite

1800 1700

60

hv bit. no. 2

Xc

ϑ(Xc)

0.6

80

1900

1600

0.4

40

1500

0.2

0.0 0

20

1400

hv bit. no. 2

1300 20

40

60

80

100

Xc

0.025

0.050

0.075

0.100

0.125

0

Time, s

Fig. 4.5 (Left) Structural evolution functions and (right) char burnout histories assigned with SNORs for (solid) hv bituminous no. 2 and (dashed) lignite for pc firing with 8% O2.

consistent with the well-established observation that lignite chars (and biomass chars) have little or no tail during the later stages of their burnout histories, whereas other coals have extended tails, as seen with the bituminous char in this example. Overall, such rapid combustion times are not surprising for 63 μm particles of any coal, as only the coarsest particles in a pc grind contribute to LOI, and furnace O2 levels are lower than the 8% in these calculations during most of the furnace transit time. The burnout histories for soot in these simulations were the same for both cases because the soot oxidation kinetics are independent of coal quality. Soot burned out considerably slower than the lignite char and at a comparable rate as the bituminous char, primarily because deactivation in the NSC rate expression is weaker than it is in CBK/E. Deactivation via thermal annealing and structural transformations are essential aspects of a char’s reactivity under pc firing conditions. Yet a conventional analysis based on nth-order char oxidation kinetics entirely omits these features. The SNOR in Eq. (4.11) is the simplest global rate expression that depicts all the essential aspects of burnout, including deactivation. Given an automated procedure to assign the required parameters from CBK/E simulations, the SNOR is easily implemented in CFD and process simulations. All that is required is a UDF that accommodates the SNOR described in this section.

4.3.6 Gasification of char and soot All the complications that arise in the analysis of char oxidation also pertain to char gasification, along with additional complexity. Char gasification is complicated by multistep surface reaction chemistry including inhibition by CO and H2, the evolution of a char’s internal pore system, and deactivation of active sites on the char surface by annealing and sintering of catalytic mineral particles throughout the conversion. Gasification rates are also mediated by the transport of the gasification agents from the free stream toward the particle center through the internal pore system, albeit with less

90

Process Chemistry of Coal Utilization

intensity than in char oxidation. In PC Coal Lab®, char gasification is modeled with a specialized version of the Carbon Burnout Kinetics Model called CBK/G. CBK/G predicts the gasification rate, the char particle temperature, and the changes in the particle diameter and density throughout conversion, given a gas temperature, radiative exchange temperature, and partial pressures of all gasification agents and inhibitors. Char reactivity is a dynamic function of heat treatment severity, based on an activation energy distribution for thermal annealing, like CBK/E for char oxidation. The theory uses a five-step surface reaction mechanism; plus transport mediation by both film and intraparticle diffusion; plus morphology changes from the Random Pore Model (RPM); plus an ash encapsulation mechanism that inhibits reactant penetration during the latest stages of char burnout. The mathematics behind these mechanisms and the energy balance to evaluate a char particle’s thermal history throughout conversion have been reported (Liu and Niksa, 2004), so only the implementation in CFD simulations is considered here. The surface reaction mechanism in CBK/G resolves gasification by steam and CO2 and hydrogasification, according to C + CO2 $ CðOÞ + CO

(4.R4)

CðOÞ ! CO

(4.R5)

C + H2 O $ CðOÞ + H2

(4.R6)

CðOÞ ! CO

(4.R7)

C + 2H2 $ CH4

(4.R8)

where the reactions are numbered for consistency with Liu and Niksa (2004) and C(O) is the oxide complex on the carbon surface. Both CO2 and steam gasification involve oxide complexes with the same nominal composition, but these complexes desorb at different rates. This variation in the desorption rates for (4.R5) and (4.R7), which arises as k5/k7 in the rate equations, depends on temperature but not pressure. This variation was invoked to describe the different saturation limits for steam and CO2 gasification of coal chars at the highest pressures. Inhibition by CO and H2 is expressed with the reversible back-reactions in (4.R4) and (4.R6). Whereas the rates of (4.R4)–(4.R7) are given by expressions of Hougen-Watson form (Liu and Niksa, 2004), the hydrogasification reaction, (4.R8), is a simple, nth-order global process, due primarily to lack of detailed kinetic information in the literature. However, hydrogasification is important only when partial pressures of H2 approach 1 MPa, which usually arises only in coal gasification at moderate temperatures and high pressures, or in co-gasification of coal and natural gas. As for char oxidation, the predicted char conversion history from CBK/G is used to specify kinetic parameters in global rate expressions that can be deployed in a CFD simulation or process simulation application. However, due to prominent roles for inhibition, deactivation, and the physical structural evolution throughout char

Support for CFD and process simulation applications

91

gasification, a single global nth-order surface oxidation reaction cannot depict the char conversion histories from CBK/G. The primary challenge in specifying char gasification rates is that almost none of the factors that actually determine these rates are uniform throughout any gasification reactor, primarily because concentrations of both gasification agents and inhibitors swing through wide ranges while particle size and char density vary continuously. Annealing attenuates the temperature dependence in the gasification kinetics. Consequently, the global rate law one uses to specify nominal gasification rates must be robust enough to depict these concentration and temperature dependences, as well as the consequences of the physical structure evolution. Contrast this situation with that for nominal devolatilization rates, where even an SFOR represents the nominal rates from FLASHCHAIN® within useful quantitative tolerances. None of the global rate laws for char gasification in the literature can depict the predictions from CBK/G. Usually, nth-order power laws or the first-order RPM are used. The temperature dependence is expressed by the activation energy parameter in a rate constant of Arrhenius form. Such an expression is adequate for the initial gasification reactivity, but cannot resolve the independent influences of the intrinsic chemistry, transport, pore evolution, and deactivation throughout an entire gasification process. Whereas an nth-order global rate law and the RPM are inadequate for simulations in gasifiers, an expansion restores their practical utility. To represent char gasification by CO2, PC Coal Lab® uses a SNOR modified for CO inhibition and structural evolution effects, as follows: nCO

RCO2 ¼ ϑ  R0CO2

ACO2  exp ð
ECO2 =RT ÞPCO22,S ¼ ϑðXC Þ  1 + KCO PCO,S

(4.15)


1 0 where RCO2 is the rate of fractional mass conversion, m1 dm dt in s ; RCO2 is the fractional mass conversion rate not subject to annealing and physical evolution effects; ϑ is a factor to account for annealing and physical evolution effects; ACO2, ECO2, and nCO2 are the preexponential factor, activation energy, and reaction order for gasification by CO2; KCO is the equilibrium constant for CO inhibition which is independent of temperature; and PCO2, S and PCO, S are the instantaneous CO2 and CO partial pressures on the particle surface. Factoring the gasification rate into separate contributions for the primary concentration and temperature dependences in R0CO2 and for the annealing and physical evolution effects in ϑ is a convenient way to expand the domain of applicability of the rate expression. ϑ represents the joint impact of the main inhibitory mechanisms that decelerate the char gasification rate with conversion, including annealing, random pore evolution, and char density changes. ϑ will be expressed as a fifth-order polynomial regression in the extent of char conversion. Similarly, the modified SNOR for gasification by H2O is

nH

AH2 O  exp ð
EH2 O =RT ÞPH22O,S RH 2 O ¼ ϑ ð XC Þ  1 + KH2 PH2 ,S O

(4.16)

92

Process Chemistry of Coal Utilization

and the SNOR for gasification by H2 is nH

RH2 ¼ ϑðXC Þ  AH2  exp ð
EH2 =RT ÞPH22,S

(4.17)

The parameters in these two rate expressions are defined in Table 4.3. For situations where all the gasification agents are present, the overall gasification rate is evaluated as the sum of the rates in Eqs. (4.15)–(4.17) and a common structural evolution factor. The global rate expression for char gasification must be implemented with an explicit account of film transport, but without any of the effectiveness factors on pore transport in CBK/G. The basic concept is that the transport of steam and CO2 across the film surrounding the fuel particle occurs in series with the consumption of steam, CO2, CO, and H2 via pore diffusion and surface reaction. Pore transport is not resolved from the surface chemistry in the global rate expression, so the series of resistances contains only film diffusion and a single, lumped surface gasification process. The global gasification rate expressions in Eqs. (4.15)–(4.17) must be implemented with the transport coefficients, product compositions, and reaction enthalpies used in the CBK/G simulations, and these parameters must be provided by the VFL for implementation in the CFD simulations. The global rates for soot gasification have the same form as those for char gasification. The only difference is that the physical evolution function for soot is much weaker than that for char because pore transport and pore system transformations are absent in soot gasification. All kinetic parameters are the same for the soots from all coal types, because soot compositions are essentially independent of coal quality. As soot contains no catalytic metals, the reactivity for soot is evaluated as that for a low volatility coal in the limit of pure carbon, which is the slowest among all coal types. Even so, soot gasification rates are comparable to chars’ because of the very small soot sizes. The CFD parameters to be specified for the gasification of char and soot are compiled in Table 4.3. They are the yields of the combustibles in char and soot, on an as-received basis; the C/H/O/N/S composition of char and the C/H/N composition of soot; and the stoichiometric coefficients for their gasification. However, for gasification, there are distinct sets of stoichiometric coefficients for gasification by H2O, CO2, and H2. The relative contributions from each gasification agent are functions of temperature, pressure, and gas composition although, generally speaking, steam gasification is fastest and hydrogasification by H2 is usually much slower than the other two gasification processes. These differences are evident in the magnitudes of the stoichiometric coefficients because the coefficients are based on a unified char gasification process as follows: 1 g
Char + νH2 O,C,H2 O H2 O + νCO2 ,C, CO2 CO2 + νH2 ,C,H2 H2 ! νH2 O,C,H2 H2 + νH2 O,C,CO CO + νH2 O,C,N2 N2 + νH2 O,C,H2 S H2 S + νCO2 ,C,CO CO + νH2 ,C,CH4 CH4

(4.18)

Support for CFD and process simulation applications

93

The primary conversion channel is steam gasification, which produces H2, CO, N2, and H2S. The H2 product is reduced to account for the hydrogen in H2S. Somewhat arbitrarily, the production of CO by the oxygen in char is added to the CO produced by steam gasification, as an expedient way to ensure that the inherent oxidizer in char contributes to the syngas composition. Also, the amount of combustibles in char that can be converted via gasification chemistry is reduced by the amount needed to react with the oxygen in char. Gasification by CO2 produces additional CO and hydrogasification produces CH4. In aggregate, all coefficients are referenced to a unit gram weight of char. The same stoichiometry is assigned for soot, except that soot has no sulfur so no H2S is produced. The stoichiometric coefficients for gasification are intended for a single, overall rate law for gasification that is evaluated as the sum of the three gasification rate expressions for H2O, CO2, and H2. Once the proportions of the various products have been specified, the heat of gasification is evaluated as a weighted sum of the values for each product formation channel. The kinetic parameters in Table 4.3 for the three global char gasification rates comprise three sets of a frequency factor, activation energy, and reaction order, two equilibrium constants, and a common set of the six coefficients in the fifth-order polynomial fit to ϑ(XC). The global rates for char and soot gasification are in units of g/cm2 s, based on the external char surface area. The material’s internal surface area does not need to be specified (although CBK/G does account for surface area variations throughout gasification with the RPM). However, as noted above, the char gasification rates from an SNOR can only be implemented by explicitly accounting for the impact of film diffusion in series with the global expression for the particle gasification rate. These kinetic parameters are adjustable constants that change with pressure, gas composition, temperature history, and coal type. Their magnitudes have no mechanistic significance whatsoever because such simple reaction rate expressions cannot possibly represent the numerous mechanisms that, in actuality, govern the kinetics of char gasification. They are usually assigned from laboratory test data. Under the best circumstances, which almost never arise in practice, the kinetic parameters in this analysis can be assigned from a database compiled for the same operating conditions as the practical application. In practice, the CFD practitioner is usually left to determine how the kinetic parameters should be adjusted to extrapolate from a calibration domain to the operating domain, and for different coal samples. It is much more efficient to use CBK/G to synthesize simulation “data” that can subsequently be analyzed for rate parameters just like one would analyze test measurements. The goal is to specify rate parameters for the modified SNORs in Eqs. (4.15)–(4.17) that are able to accurately describe the gasification rate over a complete gasification history as, for example, across an entrained flow gasifier. Consequently, the operating conditions to obtain a gasification history for the parameter assignments should be as similar as possible to the conditions in the application of interest. The procedure first evaluates the rate for a baseline extent of char conversion, then for different temperatures and different surface partial pressures of the reactant gases. Ai, Ei, Ki, and ni are assigned from the CBK/G-based rates by rearrangements of the modified SNOR expressions.

94

Process Chemistry of Coal Utilization

For example, to specify a nominal rate of CO2 gasification, select the ambient conditions of interest, then use CBK/G to predict the extents of char conversion and the gasification rates throughout a complete gasification history. From the predicted gasification history, first, evaluate the reaction rate R0CO2(1) and the surface conditions T1, 1 PCO2,S, 1PCO,S, 1PH2O,S, and 1PH2,S, near the maximum extent of char conversion, preferably one which approaches 70%. Then assign the temperature range of interest with T2, the temperature at the onset of char gasification in the CBK/G simulation, and obtain the reaction rate RCO20(2), for surface conditions T2, 1PCO2,S, 1PCO,S, 1PH2O,S, and 1PH2,S. Similarly, use CBK/G to evaluate rates RCO20(3) and RCO20(4) for 3PCO2, S ¼ 1.21PCO2,S and 4PCO,S ¼ 0.31PCO,S, respectively, where all other surface conditions are at baseline values in both cases. Values for ACO2, ECO2, nCO2, and KCO are evaluated by rearrangement of the rate expression, Eq. (4.15), as follows:

ECO2 ¼




R ln R0CO2 ð2Þ=R0CO2 ð1Þ

(4.19a)

ð1=T2
1=T1 Þ



ln R0CO2 ð3Þ=R0CO2 ð1Þ  nCO2 ¼
2 ln PCO2 , S=1 PCO2 , S KCO ¼

4

R0CO2 ð1Þ
R0CO2 ð4Þ PCO, SR0CO2 ð4Þ
1 PCO, SR0CO2 ð1Þ


 R0CO2 ð1Þ 1 + KCO 1 PCO, S
n ACO2 ¼ exp ð
ECO2 =RT1 Þ 1 PCO2 , S CO2

(4.19b)



(4.19c)

(4.19d)

This same analysis is applied to gasification with H2O and H2. The fifth-order polynomial correlation for the decay in the reaction rate with conversion is written as: ϑ ¼ a0 + a1 XC + a2 XC2 + a3 XC3 + a4 XC4 + a5 XC5

(4.18)

where XC is the extent of char conversion and ai (i ¼ 0 – 5) denote the regression coefficients. These coefficients are evaluated by fitting the product of the annealing factor, surface area factor (from the RPM), and char density factor evaluated directly from the baseline CBK/G simulation. To improve the accuracy, two separate correlations are specified for extents of char conversion above and below 10%. The test case for this comparison is based on entrained flow gasifier conditions where the gas and radiation temperatures decay continuously from 1700°C to 1450°C under 2 MPa. The fuel is hv bituminous no. 1 at a mean size of 90 μm. The gas composition varies continuously as shown in Fig. 4.6, with steam passing through an early maximum; CO and CO2 saturating by the reactor outlet; and H2 decaying monotonically. These concentration histories were imposed in this

Support for CFD and process simulation applications

95

0.7

100

bituminous no. 1

XCHAR

0.6

XSOOT 60

0.4

CO2

0.3

CO

Xc, %

Partial Pressure, MPa

80 steam

0.5

40

0.2 20 0.1 H2 0.0

0

2

4

6

8

10

0

Time, s

Fig. 4.6 Comparison of predicted conversion histories from (solid curves) CBK/G and (dashed curves) SNOR rate assignments for char and soot for a gasification case with the changing gas composition indicated by the dot-dashed curves.

simulation and do not reflect thermochemical equilibrium at any stage. Notwithstanding the wide variations in the concentrations of the gasification agents and inhibitors, the agreement from the SNOR is nearly exact throughout the entire char gasification history, and lower by no more than 4% in the soot gasification history. With this particular bituminous coal, char is more reactive than soot, albeit based on the default kinetic parameters for this coal. The performance is comparable for steam/coal ratios both higher and lower than the baseline conditions for the baseline values of all SNOR parameters. Hence, the SNOR rate parameters assigned from CBK/G results accurately depict the CBK/G-based gasification rate throughout a complete gasification history, and also handle modest extrapolations from the operating conditions used to assign the parameters. SNOR parameters assigned for diverse coal types are compiled in Table 4.5 for the same entrained flow gasifier conditions used to prepare Fig. 4.6. The CBK/G simulations were run with default kinetic parameters rather than those assigned in a one-point calibration. Even so, the predicted coal quality impacts are surprisingly complex. The frequency factors and activation energies both increase with rank which, for these operating conditions actually inverts the expected rank dependence. Instead of slower gasification reactivities for progressively higher rank, the bituminous coals are actually converted faster than both low-rank coals, albeit not by very much. Ultimate extents of conversion are 8%–9% greater with the bituminous coals. One factor responsible for this behavior is that soon after injection, the char particles reach their maximum temperatures near 1700°C. This rapidly deactivates their reactivities to well under 10% of the initial values, and the lower reactivities persist

Table 4.5 Assigned SNOR parameters for char gasification by steam and CO2 for different coals under entrained flow gasification conditions. Lignite

Subbit

hv bit 1

hv bit 2

hv bit 3

4.90E
01 14.0 1.0 9.04E
04 1.42E + 02 29.3 0.2 2.65E
02

3.17E + 00 18.7 0.9 7.98E
04 4.40E + 02 33.3 0.3 4.35E
02

2.50E + 01 24.3 0.9 2.59E
04 1.30E + 03 37.9 0.4 6.84E
02

2.63E + 01 25.1 0.9 2.49E
04 1.52E + 03 39.0 0.4 7.13E
02

1.20E + 01 23.2 0.9 4.22E
04 1.03E + 03 37.6 0.4 6.00E
02

8.76E + 00
1.04E + 01 5.58E + 00
1.32E + 00 1.39E
01
5.31E
03

8.78E + 00
1.16E + 01 6.70E + 00
1.66E + 00 1.82E
01
7.19E
03

8.68E + 00
1.24E + 01 7.45E + 00
1.90E + 00 2.12E
01
8.50E
03

8.78E + 00
1.18E + 01 6.94E + 00
1.76E + 00 1.95E
01
7.83E
03

9.07E + 00
1.13E + 01 6.47E + 00
1.62E + 00 1.79E
01
7.18E
03

1.69E + 00
1.60E
01 8.54E
03
2.35E
04 3.21E
06
1.72E
08

1.45E + 00
1.35E
01 6.95E
03
1.82E
04 2.34E
06
1.17E
08

1.27E + 00
1.14E
01 5.57E
03
1.37E
04 1.63E
06
7.56E
09

1.52E + 00
1.36E
01 6.65E
03
1.63E
04 1.94E
06
8.95E
09

1.80E + 00
1.62E
01 8.00E
03
1.99E
04 2.42E
06
1.14E
08

Rate parameters AH2O, g/atm cm2 s EH2O, kcal/mol nH2O KH2, MPa ACO2, g/atm cm2 s ECO2, kcal/mol nCO2 KCO, MPa ϑ coefficients

For XC 10% a0 a1 a2 a3 a4 a5

Support for CFD and process simulation applications

97

throughout the entire conversion history. Gasification rates for bituminous nos. 1 and 3 are essentially the same, whereas those for bit no. 2 give a faster rate, although such small differences are well within the uncertainties for default parameter assignments in CBK/G. The assigned reaction orders for steam gasification are scattered about unity, which is double the value of one-half normally assigned for nth-order char gasification kinetics. This reflects both the more complex surface gasification chemistry and the thermal annealing in CBK/G, both of which are omitted from a conventional gasification rate analysis. Inhibition of steam gasification by H2 is negligible for all coals under these operating conditions. The assigned parameters for CO2 gasification display the same tendencies, except that the order increases from 0.2 to 0.4 over this range of coal quality. Inhibition of CO2 gasification by CO is much stronger than that of steam gasification by H2, although it is still not very important in an absolute sense. The assigned structural evolution polynomials for these conditions exhibit an abrupt reactivity reduction that remains well under 0.10 while the reactor temperature continuously falls throughout the remaining transit time. The form is the same with all coals because it is dictated by the reactor temperature profile. Finally, CFD practitioners should be aware that comprehensive gasification mechanisms such as CBK/G often depict coal quality impacts that are surprisingly complex. As an illustration, Fig. 4.7 shows the char conversion histories for lignite and hv bituminous no. 2 under idealized conditions for entrained flow and fluidized bed gasifiers. Both cases have uniform gas compositions with 30% steam and 15% CO2, and uniform temperatures of 1600°C and 850°C, respectively. The solids contact times were 89 s in the fluidized bed and 10 s in the entrained flow reactor. As seen in

100 Fluidized Bed 850 °C

Entrained Flow 1600 °C

80

Xc, %

60

40

20

0

0

20

40

60

80

100

Time, s

Fig. 4.7 Predicted conversion histories from CBK/G for (solid curves) hv bit. no. 2 and (dashed curves) lignite for entrained flow and fluidized bed gasifier conditions.

98

Process Chemistry of Coal Utilization

Fig. 4.7, the conversion history develops much faster for the lignite than the bituminous in the fluidized bed, so that conversion exceeds 99% in 56 s for the lignite but is only 77.5% at the reactor outlet with the bituminous. As corroborated by numerous kinetic studies in the literature on char gasification at moderate temperatures, lignites have much faster reactivities than bituminous coals. But when CBK/G is extrapolated to entrained flow gasifier conditions with the same kinetic parameters, conversion at the reactor outlet exceeds 90% with the bituminous coal but is less than 80% with the lignite. The reason is evident in the SNOR kinetic parameters assigned for these two coals in Table 4.5 (albeit for different operating conditions). To depict the faster reactivity of lignite at moderate temperatures, activation energies in CBK/G are lower than those for bituminous coals. When lower activation energies are extrapolated to hotter temperatures, the conversion lags behind coal chars whose activation energies are higher, such as the bituminous coal. This is why the relative reactivities are opposite for the entrained flow reactor conditions in Fig. 4.7. Indeed, the effects of different kinetic parameters in complex surface reaction mechanisms like the one in CBK/G are such that the gasification reactivities can decrease, remain the same, or increase for coals of progressively higher rank, depending on the operating conditions. This complexity is also mimicked by SNORs assigned from CBK/G simulations with the analysis in this chapter.

4.3.7 An important omission The information to support CFD in the previous sections completely covers the thermophysical properties and kinetic parameters for the solid fuel phase. It is difficult to imagine a solid fuel utilization process that could not be described by the information provided by this analysis. But there is still a very important omission for any commercially important application: finite-rate kinetics for chemistry in the gas phase among volatiles, O2, H2, and H2O. In fact, the only applications for which the CFD chemistry submodel described by the kinetics in this chapter is sufficient are the typical pc furnace and the entrained-flow gasifier in which the volatiles burn at mixing limited rates in thermochemical equilibrium. In that scenario, finite-rate kinetics for the gas phase are superfluous, and all required CFD input appears in Tables 4.1 and 4.3. But in combustors, gasifiers, and pyrolyzers that operate at moderate to low temperatures, chemistry in the gas phase is neither relatively rapid (so it is not mixing limited) nor equilibrated. The challenge in these applications is that there are no simple kinetic expressions for the gas phase chemistry that accurately describe the conversion rates or products of mixtures as complex as volatiles. Chapters 5–7 describe how detailed, elementary reaction mechanisms can be incorporated into simulations of such situations, and PC Coal Lab® is routinely connected to packages that manage hundreds of elementary reactions for hydrocarbon conversion and pollutants. But these are certainly not CFD or conventional process control simulations. Moreover, elementary reaction mechanisms, as accurate as they may be, do not normally reduce to the rudimentary rate expressions in CFD chemistry submodels. Even their reduced forms are generally too complex for CFD because they incorporate too many species. So there are two

Support for CFD and process simulation applications

99

options to consider on reaction mechanisms for gas-phase chemistry, both of which involve comprehensive elementary reaction mechanisms: (1) Use an elementary reaction mechanism to develop the simplest possible quasiglobal reaction scheme that works for the application conditions of interest; and (2) Couple a VFL like PC Coal Lab® to a kinetics package such as Cantera or CHEMKIN to treat the chemistry in detail. It may be possible to specify very simple reaction schemes in some applications if the domain of operating conditions and fuel quality is not too broad. Otherwise, direct application of the elementary reaction mechanisms is often the best approach. Chapter 5 presents the theory behind this option, and Chapters 6 and 7 present numerous case studies.

4.4

Specifying representative operating conditions

There is an inherent mismatch between CFD simulations, which explicitly account for 3D geometrical effects and resolve spatial gradients, and a VFL like PC Coal Lab®, which describes transient conversion histories for individual fuel particles with no spatial resolution whatsoever. For applications involving dilute suspension phases, this mismatch is rectified by the discrete particle model (DPM) which tracks the conversion histories of a population of individual particles in a Lagrangian reference frame, then compiles coupling functions to the gas flow field from the population of results. Conceptually, the VFL should be regarded as a support system for the particle conversion submodel that feeds thermophysical and kinetic information into the DPM. Specifically, CFD practitioners should, literally, replace the particle chemistry submodel in their CFD package with a UDF that imposes the suite of global reaction rate expressions from the VFL. Although this coupling is straightforward conceptually, it is not immune from complications. One complication arises because coals are always widely distributed in size into a PSD before processing. However, it is often impractical to specify the parameters in a CFD chemistry submodel for each class of fuel particles. Some form of averaging is required to specify parameters that make the rate expressions sufficiently robust to represent all the particles that pass through a group of similar fuel injectors. As a specific example, consider a CFD furnace simulation for a T-fired furnace with fuel injectors in every corner on several levels of the lower furnace elevation. Suppose the furnace is fed with a standard utility grind of hv bituminous coal whose mean size is 50 μm. To a user preoccupied with the DPM, it may seem reasonable to specify the global rate parameters for devolatilization from a FLASHCHAIN® run that injects a 50 μm particle into gases at some very high temperature within a furnace at an even hotter temperature, because both gas temperature and radiation flux are at their maxima in the near-injector region. But this run would give an excessive particle heating rate and, therefore, overestimate the ultimate volatiles yield, swelling factor, and the frequency factor in the devolatilization rate and give a corresponding underestimate for the char yield. Among the many reasons for these discrepancies, dense fuel suspensions do not heat nearly as fast as individual particles injected into large bodies of quiescent hot gases; the temperatures of gases and particles in dense suspensions

100

Process Chemistry of Coal Utilization

increase at similar heating rates; and particles of different sizes have similar heating rates because the entire suspension heats volumetrically (Niksa, 2019b). So instead of focusing on individual particles, it would be much more accurate to average the particle histories across the suspension from a preliminary CFD simulation to assign a single mean thermal history for the entire suspension. Users could then enter into the VFL whatever temperature histories for the gas and furnace wall matched the predicted particle temperature history to the mean history for the suspension. It does not matter if the particle size or the histories of gas and wall temperature in the VFL run are different than the “true” values. When assigning CFD input parameters for devolatilization, the paramount consideration is the thermal history of the coal suspension, along with the fuel properties and pressure. Even this last assertion can sometimes be relaxed. For pressures greater than 0.5– 1.0 MPa, ultimate volatiles yields are not enhanced by rapid heating, because the elevated pressure suppresses the release of the heavier tar components that constitute the yield enhancement due to faster heating. In such applications, users have much more latitude with the particle thermal histories than they would in pc furnace simulations. Even at atmospheric pressure, uncertainties of a factor of three or so in the heating rate specified for the suspension in the VFL run would have little impact on the assigned information for the CFD simulation that pertains to devolatilization. It is also worth reiterating that, generally speaking, ultimate volatiles yields are usually much more important than devolatilization rates in CFD furnace simulations. For situations where particles of different sizes heat at different rates, one effective strategy is to run several FLASHCHAIN® cases for some number of increments in the PSD at the same operating conditions. With a SFOR, the assigned activation energies will be very similar for all sizes, and the frequency factors can be averaged with weighting by the PSD into a single, nominal value for the whole feed. If the variations in the ultimate volatiles yields are significant, the predicted values can be used to make a regression function for total yield vs size. This correlation then determines the volatiles yield used in the CFD package for each size in the simulation. As an option that accounts for variations in both ultimate yields and devolatilization rates among different sizes, various sets of devolatilization parameters can be related to ranges of particle heating rate, as in the Tabulated Devolatilization Process Model of Hashimoto et al. (2012a,b). For char oxidation kinetics in pc flames, the paramount considerations are the variations in the O2 level and temperatures of gases and walls along the combustor. In most modern furnaces, the local O2 level within the fuel suspension essentially vanishes in the near-burner region, then grows by the addition of tertiary and overfire air streams, then diminishes continuously through the upper furnace elevations. Gas and radiant (wall) temperatures exhibit maxima near the burners, then remain fairly uniform through the air injection zones before they decay along the convective passes. Such variable operating conditions should always be imposed in the VFL runs used to assign global oxidation kinetics for char conversion. Due to the impact of deactivation via thermal annealing, the magnitude and duration of the maximum particle temperature in a combustion history must be accurate; otherwise, the oxidation kinetics can be significantly biased. In fact, it is difficult to imagine that kinetic parameters

Support for CFD and process simulation applications

101

assigned for a uniform O2 level and gas temperature could possibly describe char burnout through an actual coal-fired furnace. For char oxidation in dense phase systems at moderate temperatures, particularly in fluidized bed combustors, the surface oxidation kinetics contribute more to the overall burning rates but, ironically, are less sensitive to the operating conditions. One reason is that thermal annealing is unimportant at temperatures below about 1000°C, so the char burning rates are much more uniform than those in pc flames. Another is that the particle temperatures are nearly constant, due to the fast transport coefficients in dense fluidized beds, and to the slow volumetric heat release rates in more dilute riser sections. The main consideration for fluidized beds should be an accurate value for the initial char oxidation reactivity, which always requires a one-point calibration with test data for the particular fuel sample under consideration. Unfortunately, these simplifications do not pertain to circulating fluidized bed combustors, where large particle temperature excursions and thermal annealing come into play (Niksa et al., 2017). Another important aspect for moderate temperatures is whether secondary volatiles pyrolysis will be complete in the available residence time for gases. The most direct way to assess this issue for a particular combustion system is to estimate a range of residence times and a temperature history for the gas flow, and use FLASHCHAIN®’s tar decomposition mechanism to determine the extent of secondary volatiles pyrolysis in the application. If necessary, the three-step global tar decomposition scheme can be built into the CFD chemistry submodel. For char gasification kinetics at temperatures above about 1200°C, it is best to assign kinetics in a simulation of suspension firing, which imposes the overall stoichiometric ratio for the subject gasifier and updates the gas composition with thermochemical equilibrium calculations throughout the transit time. Consequently, the concentrations of H2O, CO2, H2, and CO throughout the gasification history are bound to be reasonably close to the syngas composition along the gasifier. Temperature histories for gases and the gasifier wall must also be reasonably accurate to ensure that the assigned kinetics from the VFL will be accurate in a CFD simulation. With this approach, the greatest uncertainty in the conditions for entrained-flow gasifiers pertains to the extent of char oxidation in the near-injector region, rather than to variations in the operating conditions. In actuality, this extent of char oxidation is affected by injector design as well as by the kinetics for simultaneous oxidation of volatiles, soot, and char in a highly turbulent flow field that deposits char into a slag layer. In lieu of much more complex analyses for this situation in Chapters 6 and 7, users should vary this value in the VFL runs to assess its impact on the assigned kinetics for char gasification. If the impact is substantial, then the more thorough analysis may be justified. For char gasification kinetics in dense phase systems at moderate temperatures, the challenge is to estimate histories for the concentrations of all the gasification agents and inhibitors. Temperatures are uniform but too cool for thermochemical equilibration of the gas phase reforming chemistry. The most reliable strategy is to estimate the gas composition history from the application simulator, then enter these conditions in the VFL run that assigns char gasification kinetics. Iterate until the conditions in the VFL are reasonably close to those in the application. Unfortunately, if the finite-rate kinetics for gas-phase reforming are inaccurate, this scheme will not work well.

102

Process Chemistry of Coal Utilization

Moreover, the ambiguities surrounding the extents of secondary volatiles pyrolysis in moderate temperature combustors are even more important in moderate temperature gasifiers. First, this chemistry directly affects the syngas composition. Second, H2 is an especially effective agent for tar decomposition if its partial pressure approaches 1 MPa. This tar hydrogenation chemistry has been analyzed within the FLASHCHAIN® framework (Niksa, 2018) and is also amenable to global reaction schemes (see Chapter 9 in the first volume in this series). Although admittedly difficult to implement, realistic reaction mechanisms for all the important aspects of gasphase chemistry are essential in gasifier applications at moderate temperatures. Under all circumstances, CFD practitioners can always rely on iteration to ensure that the conditions imposed in a VFL run to assign parameters in the global rate expressions are close to those in the actual application. The sequence starts with a guesstimate for the process conditions, which are then imposed in the VFL to obtain the first set of kinetic parameters. After these parameters have been run in the application simulator, the simulation results are interrogated to specify detailed operating conditions for the second round of VFL calculations, and the second set of kinetic parameters. This iteration is closed when the operating conditions in the latest application simulation agree with those used in the previous VFL run.

4.5

Performance

The most stringent way to evaluate the CFD chemistry submodel in this chapter would be to run multiple CFD simulations on the same furnace for a diverse assortment of coals with the default parameters in the CFD application and then repeat all cases with the present chemistry submodel. Ideally, extensive field test data would be collected for operations with all coals in the simulation suite. Information like this simply does not exist because CFD simulations and field testing are too expensive to be used in hypothetical research projects like this one. And companies are not inclined to disclose it even when it is available. Moreover, it is impossible to attribute familiar performance indicators like NOX and UBC emissions to any one submodel in the CFD simulations. It is firmly established that both emissions are governed by the uniformity of the fields of temperature and O2 level which, in turn, depend on the submodels for turbulent mechanics, particle dynamics, and heat transfer as well as the chemistry submodel. Suffice to say that the chemistry submodel in this chapter is currently being implemented by dozens of utility OEMs and utility service companies in North America, Asia, Europe, the United Kingdom, and South Africa. Their feedback has noted significant improvements in burner development schedules and furnace NOX predictions after the chemistry submodels were updated with PC Coal Lab®. In lieu of the performance in commercial-scale applications, it is possible to invert this validation exercise, by imposing the operating conditions used in the VFL simulations with the comprehensive reaction mechanisms in CFD simulations with the associated global rate expressions. The chemistry submodel in this chapter was implemented in a UDF that was compiled in the CFD simulations for devolatilization only and pc firing conditions. As seen in Fig. 4.8, the CFD simulations give essentially

Support for CFD and process simulation applications 100

600 70 400

0.1 MPa, 1150 °C

60

50

0.1 MPa, 1150 °C

1.20

200

1000 800

80

600 70 400 60

3.0 MPa, 1150 °C

0 1.2

1.20

1.1

1.1 1.0

1.15

1.0

1.15

0.9

1.10

0.7 0.6 1.05

ρ, g/cm3 d/d0

d/d0

0.9 0.8

0.8

1.10

0.7 0.6 1.05

3.0 MPa, 1150 °C

0.5 1.00 0.00

0.4 0.02

0.04

0.06

Time, s

0.08

0.10

200 0 1.2

50

ρ, g/cm3

800

80

Temperature, °C

1000

90

1200

90

Temperature, °C

1200

Char Mass, AR wt. %

Char Mass, AR wt. %

100

103

1.00 0.00

0.5 0.4

0.02

0.04

0.06

0.08

0.10

Time, s

Fig. 4.8 Validation of devolatilization histories for 63 μm particles of hv bituminous no. 1 for 1150°C at (left) 0.1 and (right) 3.0 MPa from (curves) a CFD simulation with (points) results from FLASHCHAIN®.

the same devolatilization histories as the VFL under directly comparable operating conditions. For the two validation cases in Fig. 4.8, 63 μm particles of hv bituminous no. 1 were entrained at 25°C into N2 at 1150°C within a duct at the same temperature. The two cases differed in only the pressures of 0.1 and 3.0 MPa. The UDF described inert heating to the vaporization temperature of 90°C, moisture release, and complete devolatilization. Indeed, the charred mass, particle temperatures, particle sizes, and particle densities are essentially the same as those from PC Coal Lab® throughout the entire history. Only the thermophysical properties and SFOR kinetic parameters for the fuel and the operating pressures were changed in the CFD simulation to prepare these two cases. Fig. 4.9 extends the comparison to char oxidation histories, including the histories for particle temperature, particle density, scaled size, and particle mass loss. The performance of the UDF with global kinetics is comparable with CBK/E throughout. Both simulations finely resolve the ignition stage and its associated temperature surge and, especially, the reduction in the burning rate during the latter stages. Particle density diminished dramatically during devolatilization, then remained fairly uniform until the latter stages, when the greater ash density became the major contribution to the particle density. Size increased by 12% during devolatilization then diminished at a fairly uniform rate throughout the combustion history until it approached the size of a single ash particle (because this mechanism does not account for the shedding of flyash). The particle mass diminished at distinctly different rates during devolatilization and char conversion, and the burnout rate did not decelerate until

104

Process Chemistry of Coal Utilization 2400

1200

1800

Temperature

1600

1000

1400

800

1200

600

Density

800

400

600

200 0 0.0

1000

1050C, 1 atm, 8 % O2 CBKE

1.1

2000

1.0

80

Size 0.9

60

0.8 0.7

40

0.6 0.5

Mass

20

0.4

400 0.1

0.2

0.3 Time, s

0.4

0.5

100

Particle Mass, wt. %

Particle Temperature, C

1400

1.2

2200

d/d0

1050C, 1 atm, 8 % O2 CBKE

1600

Density, kg/m3

1800

0.3 0.0

0 0.1

0.2

0.3

0.4

0.5

Time, s

Fig. 4.9 Predicted combustion histories showing (left) particle temperature and density and (right) scaled size and particle mass from (curves) a CFD simulation and (points) PC Coal Lab® for a hv bituminous no. 1 injected into N2 at 1050°C with 8% O2.

the latter stages of combustion, as expected with CBK/E. The predictions with the UDF are based on the SNOR assigned by PC Coal Lab® that mimics the predictions from CBK/E for these conditions. As the SNOR does not precisely match the mass loss during the end of the combustion histories, there are minor differences in the predicted sizes and density during this stage. But these are minor flaws that could not affect any consequential aspect of combustor performance.

4.6

Summary

CFD simulations of large-scale coal utilization technologies cannot accommodate more than a handful of chemical species and, consequently, are poorly suited to address issues in which the process chemistry plays a prominent role. This is another way of saying that accurate and fully validated chemical reaction mechanisms can only be implemented in simulation platforms that accommodate dozens, if not hundreds, of chemical species. At least this is how a purist views this situation. But from a more pragmatic perspective, conventional chemistry submodels for CFD do support reaction rate expressions that are adequate for the essential aspects of solid fuel conversion, albeit only in pc furnaces, entrained flow gasifiers, and other systems that operate at the hottest temperatures. The crucial questions are, “Are those expressions able to capture the distinctive behavior of individual fuel samples?” and, if so, “How are the parameters in the available rate expressions specified to do that?” The answer provided in this chapter is an unqualified “Yes” to the first question. The answer to the second question is to use comprehensive reaction mechanisms as a stand-alone VFL that first simulates the fuel properties and operating conditions of interest and then analyzes the simulation results to specify parameters in the global expressions that closely mimic the comprehensive reaction mechanisms. Comprehensive mechanisms depict the distinctive behavior of individual fuel samples and have already been validated to predict yields and product distributions as accurately as they can be measured. So there is little doubt that the simulation “data” from the

Support for CFD and process simulation applications

105

comprehensive mechanisms accurately depict the CQ impacts. The only remaining issue is whether or not the global expressions in the CFD submodel can accurately depict the results from the comprehensive mechanisms. The findings in this chapter clearly demonstrate that all the important chemical stages of coal processing—drying, devolatilization, tar decomposition, and conversion of char and soot—can be accurately represented with global rate expressions, provided that the parameters in the global expressions are specified directly from the predictions from the comprehensive mechanisms. The analysis to specify the kinetic parameters is no more complex than that used to specify kinetic parameters from laboratory test data, and both more accurate and more convenient to implement. For drying, a moisture vaporization rate limited by heat transfer suffices, for which the particle temperature remains at the boiling point while the external heat flux provides the latent heat of vaporization until all moisture has been released. For primary devolatilization, a SFOR, DAEM, or C2SM are all potentially satisfactory, for two reasons. First, the devolatilization stage in CFD simulations is resolved with only a handful of spatial increments because particle heating rates exceed 10,000°C/s, and devolatilization is complete in tens of milliseconds. Second, the ultimate volatiles yield is much more important than the devolatilization rate, per se, because the volatiles yield partitions the fuel into components that are converted in tens of milliseconds vs char that will be converted in a few seconds. The ultimate volatiles yield, YV, is an input parameter in the SFOR and DAEM that is specified from a FLASHCHAIN® simulation, whereas the rate parameters and stoichiometric coefficient, α2, determine the ultimate volatiles yield in the C2SM. As the C2SM gives an inadvertent and uncalibrated yield enhancement for faster heating rates, either the SFOR or DAEM is strongly recommended for all CFD applications. The main advantage of the DAEM is that it accurately predicts the dynamics over three orders of magnitude in heating rate, whereas the SFOR should only be used within one order of magnitude of variation. With both these rate expressions, the impact of yield enhancements due to faster heating rates can be handled by fitting a curve to the ultimate yields from FLASHCHAIN® for the heating rate range of interest. There are two scenarios for the chemistry of primary devolatilization products. Either the entire distribution of primary volatiles can be regarded as a single species lump, or primary volatiles can be instantaneously converted into a pseudo-twocomponent mixture of noncondensable secondary pyrolysis products and soot. The first option is based on a single gaseous mixture of all the primary devolatilization products, including noncondensable gases (CO, CO2, H2O, H2, H2S, HCN, CH4, C2H4, C2H6, C3H6, and C3H8) and tar. The second option is a two-component mixture of all products of secondary volatiles pyrolysis, including noncondensable gases (CO, CO2, H2O, H2, H2S, HCN, CH4, and C2H2) and soot. By definition, the sum of the yields of secondary noncondensables and soot equals the yield of the whole primary volatiles. In large coal flames and entrained flow gasifiers, the conversion of primary volatiles into secondary gases plus soot is nearly instantaneous, so the two-component mixture option accurately estimates the gaseous fuel components that actually burn in near-burner flame zones. Chemical kinetics for tar decomposition into soot are

106

Process Chemistry of Coal Utilization

superfluous. However, for coal processing at temperatures cooler than 800°C or so, the time scales for tar decomposition become comparable to those for primary devolatilization, and oils and PAH become the major decomposition products, instead of soot. For such conditions, finite-rate kinetics for tar decomposition should be incorporated into the CFD chemistry submodel, along with detailed kinetics for chemistry among species in the gas phase. The sample-to-sample variability in ultimate volatiles yields is particularly significant for ranks through hv bituminous, above which the yields diminish rapidly for low volatility coals. Differences among the volatiles compositions are even larger. For whole primary volatiles, the carbon per mole of volatiles is up to 60% greater for hv bituminous volatiles, and H/mol is up to one-third greater, whereas O/mol is lower by up to 40%. Consequently, the O2 requirement for stoichiometric volatiles combustion is up to 75% greater for bituminous than for lignite. Even among different hv bituminous coals, the variation in the O2 stoichiometry is substantial. Similarly, the heats of combustion increase by up to 50% for primary volatiles of progressively higher rank. For instantaneous conversion of primary volatiles into secondary noncondensable volatiles plus soot, the predicted soot yields are markedly lower for most, but not all, low-rank coals compared to hv bituminous. The molecular weights of noncondensables diminish for progressively higher rank, due primarily to the abundance of CO2 in the volatiles from low-rank coals. Volatile compositions of the secondary noncondensables are completely different than those of primary volatiles, due to the reforming of GHCs into CH4 and C2H2 and the production of CO and H2 during the decomposition of tar into soot. Carbon-contents diminish for progressively higher rank, opposite to the tendency in primary volatiles. They are only one-tenth the levels in primary volatiles, while both H- and O-contents are halved in the secondary noncondensables. The N-contents are also diminished because about one-third of tar-N is incorporated into soot. The O2 requirement for stoichiometric volatiles combustion increases for progressively higher rank, although the strength of this tendency is markedly weaker than for primary volatiles. Heats of combustion of secondary noncondensables vary by 10% or less among these coals. The soot compositions with all coals are nearly the same at 98.5% C, 1.0% H, and 0.5% N. Unfortunately, none of the global rate expressions in the literature are adequate for char conversion via either oxidation or gasification. These rate expressions can depict how variations in temperature and reactant concentrations affect the conversion rate within a narrow domain of operating conditions. But they cannot be extrapolated to other conditions, especially more severe conditions, because they do not account for thermal annealing, continuous char density changes, physical structure evolution, and, in gasification, inhibition by CO and H2. Furthermore, these global rate expressions also cannot depict deceleration in the char conversion rates during the latter stages of conversion due to annealing and ash encapsulation. The expressions for nth-order or first-order char conversion rates are simply no substitute for the more complex surface chemistry, particularly when the partial pressures of the reactants vary over wide ranges, as they do in pc flames and gasifiers.

Support for CFD and process simulation applications

107

Fortunately, the utility of nth-order rate laws can be recovered by factoring the char conversion rate into a contribution for physical structure evolution and another for the chemistry and pore transport within a particle. The term “physical structure evolution” is used broadly to comprise thermal annealing, changes to pore morphology, density variations, and ash encapsulation; i.e., all factors that diminish a char conversion rate throughout combustion and gasification. For char oxidation, the physical structure evolution factor can be combined with a conventional nth-order rate law, and soot oxidation can be evaluated with NSC kinetics. For char and soot gasification, the physical structure evolution factor is combined with an nth-order global rate with a term for inhibition in the denominator. Three such expressions represent gasification by steam, CO2, and H2, where the rate for hydrogasification omits the inhibition term. As for primary devolatilization, char and soot conversion is amenable to the strategy that first simulates the fuel properties and operating conditions of interest with CBK/E for char oxidation and with CBK/G for gasification of soot and char and then analyzes the simulation results to specify parameters in the global expressions that closely mimic the comprehensive reaction mechanisms. For char oxidation, the frequency factors diminish and the activation energies are greater for coals of progressively higher rank, a situation which depicts the established tendency for slowerburning rates for chars of progressively higher rank. The assigned reaction orders are scattered about unity, which is double the value of one-half normally assigned for nth-order char oxidation kinetics. This reflects the more complex surface oxidation chemistry and thermal annealing in CBK/E, both of which are omitted from a conventional burning rate analysis. The assigned structural evolution factors are fifth-order polynomials in the extent of char conversion that decay continuously while char temperatures surge immediately after ignition, and may or may not saturate to a stable, low value during the latest stages of burnout. For char gasification, the reaction orders for steam gasification tend to be greater than those for CO2 gasification. The kinetic parameters assigned for SNORs are often mixed, with no clear tendency in rank. This may seem surprising because gasification reactivities decrease for progressively higher rank coals at moderate temperatures. This tendency is depicted with activation energies that increase for progressively higher rank. However, when lower activation energies are extrapolated to hotter temperatures, the conversion lags behind coal chars whose activation energies are higher. Indeed, the effects of different kinetic parameters in complex surface reaction mechanisms like the one in CBK/G are such that the gasification reactivities can decrease, remain the same, or increase for coals of progressively higher rank, depending on the operating conditions. This complexity will also arise in SNORs assigned from CBK/G simulations with the analysis in this chapter. For both oxidation and gasification, the discrepancies among the CBK-based conversion histories and those based on the associated SNORs are inconsequential. This covers cases where char oxidation is extinguished to a slow burn state soon after a burning rate becomes film diffusion-limited and gasification environments where the pressures of all gasification agents and inhibitors vary through wide ranges during the reactor transit time.

108

Process Chemistry of Coal Utilization

References Hashimoto N, Kurose R, Hwang SM, Tsuji H, Shirai H. A numerical simulation of pulverized coal combustion employing a tabulated-devolatilization-process model (TDP model). Combust Flame 2012a;159:353–66. Hashimoto N, Kurose R, Shirai H. Numerical simulation of pulverized coal jet flame employing the TDP model. Fuel 2012b;97:277–87. Liu G-S, Niksa S. Coal conversion submodels for design applications at elevated pressures. Part II. Char gasification. Prog Energy Combust Sci 2004;30(6):697–717. Merrick D. Mathematical models of the thermal decomposition of coal. 2. Specific heats and heats of reaction. Fuel 1983;62:546. Niksa S. Flashchain theory for rapid coal devolatilization kinetics. 11. Tar hydroconversion during hydrogasification of any coal. Energy Fuels 2018;32:7569–84. Niksa S. Simulating volatiles conversion in dense burning coal suspensions. Part 1. Validation of reaction mechanisms. Fuel 2019a;252:821–31. Niksa S. Simulating volatiles conversion in dense burning coal suspensions. Part 2. Extrapolations to commercial p.f. firing conditions. Fuel 2019b;252:832–40. Niksa S, Liu G-S, Hurt RH. Coal conversion submodels for design applications at elevated pressures. Part I. Devolatilization and char oxidation. Prog Energy Combust Sci 2003;29:425– 77. Niksa S, Sakurai Y, Fujiwara N. Predicting the conversion efficiencies of any coal type in CFBCs. Energy Fuels 2017;31:4507–19. Park C, Appleton JP. Shock-tube measurements of soot oxidation rates. Combust Flame 1973;20:369–79. Patil S, Cooper J, Orsino S, Meadows J, Valdes R, Laster WR. Investigation of single-jet combustor near lean blowout conditions using flamelet-generated manifold combustion model and detailed chemistry. J Eng Gas Turbines & Power 2016;138. Article number 121503. Strugala A. Empirical formulae for calculating real density and total pore volume of hard coals. Fuel 1994;73(11):1781–5.

Simulations with detailed chemistry

5

Nomenclature aSOOT ASOOT (C/O)j Deff, O2 %Ei fASH FCO FC 0 FC j FE FEj FNO FP j FS ∞ ki M C0 mchons mi Mi Δmi,k m O2 MS 0 mvol N Nf nP nRXN pO2 QO2 s SA SB SD SH SO t ti T̄ g, i

specific surface area of soot, cm2/g total surface area of soot in a CSTR, cm2 carbon to oxygen ratio in the products from CSTR j turbulent diffusivity of O2 weight percentage of element E in char or soot, daf wt.% initial mass fraction of ash in coal fraction of char-C oxidized into CO mass flowrate of char based on its ultimate yield, g/s char flowrate into CSTR j fraction of an airstream entrained into the process flow to time t gas entrainment increment into CSTR j fraction of char-N oxidized into NO throughout char burnout mass flowrate of a primary gas stream into CSTR j mass flowrate of soot based on its ultimate yield, g/s Arrhenius rate constant in NSC kinetics, i ¼ A, B, Z char molecular weight based on its elemental composition, g/mol local mass fraction of combustibles in a CFD simulation local mass fraction of species i in a CFD simulation molecular weight of species i change in the mass of species i from gasification reaction k fractional entrainment rate of O2 into a region, s1 soot molecular weight based on its elemental composition, g/mol local mass fraction of volatiles in a CFD simulation number of CSTRs in a CSTR-series number of fluid-particle trajectories across a flow region number of particles in a computational cell in a CFD simulation number of reactions for oxidation and gasification for char or soot partial pressure of O2 in NSC kinetics entrainment flux of O2 across the boundary of a flow region boundary of a flow region reactive sites for soot oxidation bulk carbon sites on soot nearly unreactive sites for soot oxidation adsorbed hydrogen on soot adsorbed oxygen on soot time, s mean gas residence time in CSTR i mean gas temperature for the ith time increment into a distinctive flow region

Process Chemistry of Coal Utilization. https://doi.org/10.1016/B978-0-323-89959-8.00005-7 Copyright © 2022 Elsevier Ltd. All rights reserved.

110

Tg,i,j u0,j V Vcell Vreactor xE ΔXjC,O2 ΔXjC,k ΔXjS,O2 ΔXjS,k yA 0 YC 0 yE, i j y P, i j

Process Chemistry of Coal Utilization

gas temperature for trajectory j in time increment i across a flow region mean gas velocity of trajectory j across a flow region volume of a CSTR computational cell volume in a CFD simulation CSTR volume, cm3 mole fraction of element E in char extent of char burnout in CSTR j extent of char conversion by gasification reaction k in CSTR j extent of soot burnout in CSTR j extent of soot conversion by gasification reaction k in CSTR j mass fraction of ash in char from primary devolatilization, wt.% yield of char from primary devolatilization, daf wt.% mass fraction of species i in gas entrainment increment into CSTR j mass fraction of species i in the primary gas stream into CSTR j

Greek symbols βM χ χA χO ρ ρdaf ρtotal τM νC,i,k νC,i,O2 νS,i,k νS, i,O2 ωi ωNSC ωSOOT ψ

empirical mixing constant in an entrainment rate law ratio of rates in NSC kinetics ratio of rates for reactive open sites in soot oxidation kinetics ratio of rates for adsorbed oxygen sites in soot oxidation kinetics gas density, kg/m3 particle density of daf coal, kg/m3 total density of material in a CSTR, kg/m3 time lag before entrainment begins from a stream into a flow region stoichiometry for species i in char gasification reaction k, mol/mol of organics in char stoichiometry for species i in char oxidation, mol/mol of combustibles in char stoichiometry for species i in soot gasification reaction k, mol/mol of organics in soot stoichiometry for species i in soot oxidation, mol/mol of combustibles in soot net species production rate from homogeneous chemistry, moles/s soot burning rate from NSC kinetics, g/cm2 s soot burning rate from the surface oxidation mechanism, g/cm2 s conserved scalar for all combustibles in a flow regardless of phase or extent of conversion

Subscripts C CO2 E i k H2 H2O O2 P S vol

char pertaining to CO2 gasification entrainment increment into a CSTR molecular species i char conversion reaction k pertaining to hydrogasification pertaining to H2O gasification pertaining to oxidation primary gas stream into a CSTR soot volatiles

Simulations with detailed chemistry

111

The previous chapter framed a conspicuous gap between the very limited chemistry in CFD simulations and the essential role that chemistry plays in the mechanisms of pollutant formation in most coal utilization technologies. This gap could not persist because the imperatives on pollutant controls have already been driving the development of coal utilization technology for decades. The gap was bridged around the turn of the century, but not by expanding the capabilities of CFD. Instead, a completely different simulation method that accommodates full elementary reaction mechanisms for chemistry in the gas phase was applied to coal utilization technologies. The method handles hundreds of species and reactions but also requires radical simplifications to other physical aspects of the reaction systems, particularly the effects of turbulence. This chapter lays the foundation for simulations of coal utilization technologies that combine comprehensive reaction mechanisms for coal decomposition, combustion, and gasification with fully validated elementary reaction mechanisms for homogeneous chemistry. It first reviews two alternative technical approaches in the next section. The rest of the chapter develops the protocols and theoretical relations for simulations based on equivalent chemical reactor networks. Chapters 6 and 7 present a variety of applications for pilot- and commercial-scale coal utilization technologies.

5.1

Laminar flamelet models vs chemical reactor networks

When one envisions a truly comprehensive simulation of a reacting coal suspension, the ultimate result is usually a 3D field that indicates temperature, velocity, and all species concentrations at every point within the boundaries of a furnace or gasifier. From these fields, conversion rates of all fuel components and production rates of all intermediates, products, and pollutants can be calculated directly. In fact, any engineering aspect of the thermal, fluid mechanical, and chemical performance could be determined from these fields, albeit only in principle. Problem is, the fidelity of the physicochemical mechanisms in the simulations determines the accuracy of the predicted fields. Now and for the foreseeable future, no simulation platform can accommodate sufficiently robust physical mechanisms along with detailed chemical reaction mechanisms to deliver accurate fields throughout reacting coal suspensions. To date, the only means to incorporate detailed chemical reaction mechanisms into simulations of coal utilization technology is to analyze the physical transformations and the chemistry in sequential calculation sweeps, rather than in a single, comprehensive simulation. The order of the sequence—whether chemistry is treated in the first or second calculation pass—differentiates the two major approaches developed for coal conversion systems. These approaches are called laminar flamelet modeling and chemical reactor networks. In flamelet modeling (Wen et al., 2020a,b), detailed chemical reaction mechanisms are used to predict the conversion rates and heat release rates across 1D laminar flames of coal volatiles. These calculations are repeated numerous times to compile a library of all reaction products from the

112

Process Chemistry of Coal Utilization

flamelets, including pollutants, for the domains of stoichiometry and temperature throughout the subject reacting coal suspension. Then the actual physical system is simulated with detailed fluid mechanics and particle dynamics, usually in so-called large-eddy simulations (LES). At points in the flowfield that meet ignition criteria on stoichiometric ratio and temperature, the volatiles are converted at the rates for comparable conditions in the flamelet library. In this manner, the simulations with detailed chemistry are de-coupled from the LES, in so far as a look-up tabulation of volatiles conversion rates replaces an explicit integration among the fluid mechanics and chemistry in the LES simulation. With chemical reactor networks, the chemistry is treated in the second calculation pass rather than the first. The first pass is a conventional CFD simulation that provides temperature and flow fields throughout the subject utilization system. The species concentration field is largely ignored because it reflects the rudimentary chemistry in CFD chemistry submodels. The temperature and flow fields are interrogated to identify an “equivalent” reactor network for the flowfield; that is, an arrangement of perfectly mixed and completely unmixed reactors that mimics the entrainment rates and transit times across distinctive regions of the flowfield. Given this equivalent reactor network, the chemistry within each reactor can be simulated with reaction mechanisms of any complexity, due to the idealized mixing attributes of each reactor element. In this manner, the chemistry is de-coupled from the limited number of species in the CFD simulation, and the equations for the detailed reaction mechanisms are never integrated with the equations for the flow and thermal fields. Laminar flamelet models are not considered further because they are still under intensive development, especially in Germany, Japan, and China. Current research is validating results for lab-scale flames in the round jet configuration, including NOX production mechanisms (Luo et al., 2019; Zhao et al., 2020) and the impacts of various assumptions, strategies, and implementation formats on the predicted flame characteristics (Watanabe et al., 2017; Luo et al., 2018; McConnell and Sutherland, 2020). Larger coal flames have also been analyzed (Wen et al., 2016; Reith et al., 2017). Applications in commercial systems are on the distant horizon so now is the time to chart the progression of this potentially important method.

5.2

Historical development of chemical reactor networks

It is important to recognize that chemical reactor networks have no means at all to resolve any of the physics associated with mixing phenomena in turbulent reacting flowfields. If one imagines a spectrum of approaches to analyze turbulent mixing, direct numerical simulations (DNS) would appear on the extreme limit for complete resolution of the fluid mechanics at all length scales down to molecular mixing via Brownian diffusion. LES would be at an intermediate position as it resolves mixing to some threshold length scale but omits the phenomena at smaller scales. CFD would be further away because it uses eddy dissipation concepts. Chemical reactor networks would be at the opposite limit, because they use flow and mixing models that completely ignore the turbulence and its associated fluctuations, and represent

Simulations with detailed chemistry

113

only the bulk flow patterns. Consequently, all turbulence/chemistry interactions are also omitted. Modeling approaches based on bulk flow patterns were developed during the 60s, 70s, and early 80s in the academic chemical engineering community, and thorough analyses are available for the performance of simple reaction mechanisms across simple networks. The simplest systems are treated in textbooks on chemical reaction engineering (e.g., Fogler, 1993), and Shinnar’s (1987) collection of examples is especially comprehensive. Applications with complete elementary reaction mechanisms came later as an outgrowth of the validation of detailed reaction mechanisms for laminar flames throughout the 1980s. Indeed, computer packages such as Cantera and CHEMKIN automatically solve an elementary reaction mechanism of arbitrary complexity across any plug flow reactor (PFR) and continuously stirred tank reactor (CSTR) and now support networks of CSTRs and PFRs. By the mid-1990s, all the capabilities were in place to develop reactor networks for pc furnaces. One of the early applications used a reactor network to simulate the mixing and fluid mechanics in a single coal burner with swirled secondary and tertiary air (Pedersen et al., 1998a). The furnace schematic and its associated reactor network appear in Fig. 5.1. The near-burner zone is modeled as a single CSTR. It delivers gases into both a jet expansion zone and an external recirculation zone. The recirculation zone is represented with another CSTR, whereas the expansion zone is represented as a PFR with distributed feedstreams for entrainment of recirculated gas into the

CSTR Jet impact (Xp)

External Recirculation Zone

Air inlet Coal inlet

CSTR

PFR

Near Burner Zone

a mn

CSTR

Jet Expansion Zone Entrainment of fluid

t=0.50 s, a=0.64 T=1600 K

Down Stream Zone

Supply of staged air

ERZ CSTR

PFR

NBZ t=0.033 s T=1750 K

CSTR

3 CSTR

DSZ

DSZ

JEZ t=0.19 s T=1750 K

t=0.81 s t=2.43 s T=1750-1612 K T=1612-1200 K

Fig. 5.1 (Top) Cylindrical coal flame and (bottom) its equivalent reactor network. Reproduced from Pedersen LS, Glarborg P, Dam-Johansen K, Hepburn P, Hesselmann G. A chemical engineering model for predicting NO emissions and burnout from pulverized coal flames. Combust Sci Technol 1998a;132:251–314 with permission from Taylor & Francis.

114

Process Chemistry of Coal Utilization

primary jet. The flow passes from the expansion zone into a postflame zone, which is modeled with four CSTRs in series. The residence times and reactor volumes in the various zones were specified to match measured values for a swirled coal flame in a tubular furnace operated by Mitsui Babcock, UK. Temperatures in each flow component were also specified directly from measurements, so energy balances were omitted. Since the characteristic time for devolatilization is shorter than the residence time in the first CSTR, no finite devolatilization rates were needed in the reactor simulation. Volatiles yield and product compositions were either measured or estimated. Volatiles were assumed to consist of only CO and H2 diluted by H2O, CO2, and N2. The only N-species in volatiles mixtures was HCN. Consequently, no C1/C2 chemistry was included in the homogeneous reaction mechanism and, therefore, reburning effects were also omitted in the nitrogen chemistry. The model also featured a detailed submechanism for soot oxidation that accounted for radical recombinations, oxidation by O and OH, and NO reduction on soot (presented in Section 5.4.2). Char oxidation was based on the conventional expression for the burning rate in the limit of shrinking particles. Predicted concentrations of the N-species for ranges of the flame stoichiometry were qualitatively similar to measurements in lab-scale flames in showing that total fixed nitrogen passes through a minimum for progressively greater SR-values due to the elimination of NH3 and HCN and the production of NO (cf. Fig. 8-9 in the first volume of this series). This same group teamed up with researchers at Danish and British utility services companies to use reactor networks to predict temperature profiles and UBC emissions from pilot- and full-scale furnaces burning three coal types (Van der Lans et al., 1998), although the evaluation of the predicted UBC emissions was unsatisfactory. To analyze NO formation in a turbulent natural gas diffusion flame, Ehrhardt et al. (1998) first simulated the flowfield with conventional CFD to assign the axial velocity field, local turbulent diffusivity, and temperature field throughout the combustor. Then a much coarser grid was overlaid on the original CFD grid based on the streamlines, and the volumes between streamlines were subdivided into a PFR network with the same axial throughput and lateral diffusion fluxes as in the CFD calculations. That is, the transport rates in the CFD simulations were explicitly imposed in the equivalent reactor network. Back mixing was neglected in the postprocessing calculations to eliminate all elliptic terms (even though the CFD simulation was elliptic because of radiation transfer). About 1000 PFRs represented the transport in the simulations with full chemistry, and each one added 25 s to the simulation time. The main drawback to this approach is its restriction to hyperbolic flows and unsuitability for elliptic flowfields. A group at the Italian utility ENEL was the first to use equivalent networks of idealized reactors to postprocess a CFD simulation of a full-scale furnace with detailed chemistry (Benedetto et al., 1997). Their approach for full-scale boilers is outlined in Fig. 5.2. The flowfield from a 3D CFD boiler simulation was first analyzed for regions of relatively uniform temperature and moderate velocity gradients. The maps of these regions were then translated into an equivalent network of reactors whose volumes sum to the furnace volume. Although details are sketchy in the published report, it

Simulations with detailed chemistry

115

3D - CFD

Flow field

Temperature

Concentration Stoichiometry

Analysis of different combustion zones

Stoichiometry Temperature Concentration

Correlation Analysis Volume reactor definition

Network of reactors

Reactor type Mass flow rates

Network dinamic response Residence time distribution NOx esperimental data

Kinetic model NOx evalutation

Network check

Fig. 5.2 Scheme to analyze CFD simulations for dynamics to automatically devise reactor networks for postprocessing with detailed chemical reaction mechanisms. Reproduced from Benedetto D, Pasini S, Falcitelli C, La Marca C, Tognotti L. NOX emission prediction from 3-D complete modeling to reactor network analysis. Combust Sci Technol 2000;153:279–94 with permission from Taylor & Francis.

appears that the method of Ehrhardt et al. was also applied by Benedetto et al. (1997). Once the reactor network was specified, its dynamic behavior was automatically determined to identify the residence time distributions (RTDs) for different furnace zones and all inlet streams. Flows and volumes of the reactors were fine-tuned to match the furnace RTD based on the CFD simulation. Full chemistry was then simulated across the reactor network. According to the first published report from this group, this approach worked well for reburning and air-staged burnout zones in a coal-fired utility furnace, but performance in the near-burner zone was unsatisfactory (Benedetto et al., 1997).

116

Process Chemistry of Coal Utilization

A second application predicted NOX emissions from a 75 MW gas combustor, where the fields from the CFD were first subdivided into zones based on bulk flow patterns (Antifora et al., 1999). Then, the SR values and temperatures in the cells in each zone were classified into three groups: rich (with SR 1.3). The cells in the rich and lean groups were aggregated into PFRs, whose volumes and flowrates were specified from the CFD simulation. The cells in the mixing group were aggregated into CSTRs. Reactor temperatures were assigned as weighted average values over the cells in each reactor group. The reactor network for the gas combustor contained 13-elements so that each distinct zone in the furnace (hopper, first burner row, second burner row, reburning zone, and OFA zone) was represented with only two or three reactors. Despite the coarse resolution, the predicted NOX emissions were generally within 20 ppm of the measured values (of approximately 200 ppm) for two baseline cases, two OFA cases, and one case with OFA and flue gas recirculation (Antifora et al., 1999). The third application from the ENEL group predicted NOX emissions from a 320 MW oil-fired furnace equipped with low-NOX burners in an opposed-wall firing configuration (Benedetto et al., 2000). In this analysis, the researchers exploited a relation between temperature and local SR in the CFD fields that could be split into two components. One displays the familiar functional dependence of the adiabatic flame temperature on stoichiometric ratio, with a maximum just to the rich side of the stoichiometric point (SR ¼ 1). The second component was a band with a nominal equivalence ratio of 0.7 that extended from very low temperatures to the flame temperature and was associated with the injection and atomization of the primary fuel streams. The reactor network for the lower half of the furnace had only eleven elements. As with the analysis of the gas combustor, the predicted NOX emissions were in good agreement with measured levels in the flue gas. From 2000 onward, the general approach for equivalent reactor networks has been implemented by numerous groups around the world and covered a wide variety of coal processing applications. However, the same variety of approaches for assigning an equivalent network from a CFD simulation seen in the earliest applications also pertains to more recent examples. Simply put, there is no consistent, validated method to assign an equivalent network. Here, the focus shifts toward the author’s approach because this was consistently implemented in the bulk of the application case studies in Chapters 6 and 7. It is called ChemNet CFD postprocessing.

5.3

ChemNet CFD postprocessing

This section describes a protocol to specify an equivalent reactor network for reacting coal flows that can be simulated with CFD. That qualification excludes processing in bubbling fluidized beds like atmospheric fluidized-bed combustors (AFBCs) and pressurized fluidized-bed combustors (PFBCs), although hybrid approaches that combine CFD with the empirical hydrodynamics for portions of fluidized systems are feasible, as already demonstrated for CFBCs (Niksa et al., 2017). Fixed bed reactors would not be analyzed with this approach either. It is well-suited for pc furnaces

Simulations with detailed chemistry

117

and entrained flow gasifiers. There are no restrictions on the scale of the application, although the specialized labor requirements may be formidable for applications at a commercial scale. Succeeding sections first describe the major elements in the analysis, then a formal implementation protocol, then the technical prerequisites. Finally, the mathematics to incorporate detailed reaction mechanisms for coal decomposition and char conversion with elementary reaction mechanisms for the gas phase chemistry are developed for nonreactive atmospheres, furnaces, and gasifiers.

5.3.1 Overview The network is “equivalent” to the CFD flowfield in so far as it represents the bulk flow patterns in the flow. So what, then, are bulk flow patterns? Bulk flow patterns are determined by the field of time-averaged velocities. Even though the average velocity field excludes turbulence, it does characterize zones of intense mixing, bypassing, recirculation, dead zones, and plug flow. In complex flows with multiple inlet streams, the RTDs of all regions emanating from injectors or burners must be specified. Once the RTDs have been specified, they are regenerated with a network of idealized reactor elements. Networks contain only two reactor types: (1) CSTRs whose RTD is an exponential decay; and (2) PFRs whose RTD is a Dirac deltafunction at the uniform residence time for all fluid elements. Numerous CSTRs in series are equivalent to a PFR, because back mixing and radial dispersion are omitted in the PFR. Consequently, the CSTR is the archetypal reactor element. The “equivalence” in an equivalent reactor network covers transit times, temperatures, and local equivalence ratios. The RTDs in the major flow structures are the same in the CFD flowfield and in the section of the reactor network that represents the flow region under consideration. Mean gas temperature histories and the effective ambient temperature for radiant heat transfer are also the same. The entrainment rates of surrounding fluid into a particular flow region are evaluated directly from the CFD simulation. To the extent that the RTD, thermal history, and entrainment rates are similar in the CFD flowfield and reactor network, the chemical kinetics evaluated in the network represents the chemistry in the CFD flowfield, albeit only in a steady, spatially averaged way. From a practical perspective, it is only possible to implement ChemNet after the CFD flowfield has first been subdivided into regions. The regions are the macroscopic boundaries for the chemical structure of the flowfield. As such, each region sustains a distinctive set of chemical reaction mechanisms. Regions are usually much more extensive than any distinct flow structures. For example, the core formed by the primary coal jet within a dual register burner is a region, because the very high loadings of particles and soot in this region will significantly bias the chemical reaction rates in the gas phase, especially the N-conversion mechanisms. Mixing layers formed by simultaneous entrainment of fuel-rich fluid into secondary or tertiary air streams are also regions, because the transverse temperature profiles along the direction of mixing exhibit similar maximum values along the entire layer. The portion of an OFA jet remaining to be mixed with a process stream is another region, because the absence of fuel essentially eliminates all chemistry.

118

Process Chemistry of Coal Utilization

The first step in developing an equivalent reactor network is to subdivide the CFD flowfield into its distinctive regions. In addition to distinctive chemistry, regions must have operating conditions that can be expressed as functions of transit time because, by definition, a network of idealized reactors reduces all spatial variations to a time dependence. This condition imposes several constraints on how regions are defined, as follows. The flowfield determines the residence times of all fluid particles moving across a particular region. As regions are generally fed by multiple streams of grossly different compositions, the flowfields within regions are rarely one-dimensional. Multidimensional flowfields determine RTDs, rather than a nominal residence time. For example, suppose that a region was defined as a round turbulent jet emanating from a cylindrical injector. Most of the fluid remains near the jet axis and travels far downstream from the injector within this region. This fluid has the longest residence time. But some of the fluid has a sufficiently fast radial velocity component to quickly move off the axis and cross the boundary of this region. Such fluid has a much shorter residence time. By tracking many fluid particles over the injector cross-section, an RTD can be determined for the region that accounts for the multidimensional character of the flowfield, without restricting the flowfield to a single coordinate. The RTDs for all regions in the CFD simulation must be matched in the equivalent reactor network to depict the impact of the multidimensional flow character on the chemical kinetics. This is implemented by recasting the CFD flowfield into a Lagrangian field of individual trajectories in time for both fluid particles and actual fuel particles. The most versatile way to match the RTDs is to represent the operating conditions in each region by an assembly of CSTRs and PFRs. These two reactors represent the extreme extents of back mixing of products with reactants, in that CSTRs are completely back mixed whereas PFRs have no back mixing. This feature is responsible for their characteristic RTDs as well. A CSTR RTD is an exponential decay and, therefore, as broad as possible. The PFR RTD is a Dirac delta function with no dispersion whatsoever. Most important, the RTDs of CSTRs-in-series can be varied continuously between these limiting forms simply by varying the number of CSTRs in the series. In practice, only series of CSTRs are used because the RTD of a PFR equals that of a CSTR-series in the limit of a large number of reactors. Regions whose RTDs did not fall within this range have been identified in furnaces, but could nevertheless be represented by the RTD of a more complicated reactor assembly, such as a CSTRseries in parallel with a PFR (cf. Fig. 5.5). All the operating conditions that affect chemical kinetics must also be recast into functions of a common time coordinate. The gas temperature field within each region must be reducible to a thermal history; i.e., an average temperature as a function of time. In principle, the profile could be expressed in terms of one spatial coordinate or in terms of a time coordinate. The time coordinate is preferable because a thermal history maps directly onto the average residence time profile along a series of CSTRs. This stipulation is potentially confusing to implement because it certainly does not imply that the gas temperature field within each region must be one-dimensional. Rather, it means that the gas temperature field must be amenable to meaningful averaging, whereby each fluid particle is subjected to a similar thermal history, regardless of its particular residence time in the RTD.

Simulations with detailed chemistry

119

To illustrate this point further, consider a 2D, axisymmetric, laminar diffusion (Burke-Schumann) flame. This flame consists of a relatively cool core of fuel, surrounded by air at ambient temperature. The interface between these two regions is a reacting surface fed by fuel from one side and by air from the other. The interface also determines the locus of maximum temperatures for the entire flame, so the gas temperature field is definitely not one-dimensional. Nevertheless, each fluid particle that moves from the fuel core into the flame surface is rapidly heated to the flame temperature, then cooled as it penetrates into the air stream. The crucial point is that the imposed thermal history is essentially independent of position on the flame surface. Whether the fluid leaves the fuel core immediately after leaving the burner or from the streamline on the flame axis into the flame tip, essentially the same thermal history is imposed: It rapidly increases from the low value in the fuel core, passes through the maximum value at the flame surface, then diminishes to the low value in the air stream. In such situations, fluid particles are tracked to compile a population of thermal histories for all trajectories represented by the RTD. Temperatures of the population in a normalized time coordinate are then averaged to evaluate an average gas temperature history for the region under consideration. Once the average temperature history has been specified, it is rendered in isothermal steps to each of the reactors in the CSTR-series for the subject region, because each CSTR must be isothermal. Provided that many CSTRs are used to represent the region, there is little uncertainty introduced by rendering the average thermal history into discrete form. In addition to the gas temperature history, two additional thermal histories must be specified. Both pertain to the particle phase. One is an effective ambient (wall) temperature for radiation transfer, and the other is a mean thermal history for the fuel suspension immediately after injection, to calculate the rates and yields for devolatilization. In principle, the radiation analysis in the CFD simulation was already used to evaluate the radiation flux to the particle along each particle trajectory in the CFD simulation. This flux could be used to directly evaluate an effective ambient temperature. In practice, this would entail a deep interrogation of the CFD simulation that is hard to justify, because downstream of near-burner regions the effective ambient temperature is usually much cooler than the particle temperature, which renders it negligible for radiant heat transfer. In practice, effective ambient temperatures are specified as average values over various sections of the surroundings, as explained further below. Effective ambient temperature histories are also implemented in discrete forms across the CSTR-series. The second required thermal history for the particle phase is only needed at fuel injectors. To run FLASHCHAIN® for primary devolatilization, a representative thermal history must be specified for the entire suspension from each injector or burner. Usually, this is not ambiguous because, for the relatively high mass loadings in commercial burners, the suspension and primary air streams have very similar temperature histories before ignition, and these histories are insensitive to particle size. A thermal history for devolatilization is assigned as a spatial average of the histories for all particle trajectories over the cross-section of the suspension at various axial positions from each injector. The average usually covers 80–100 ms, although devolatilization is usually complete in shorter periods. The thermal history for devolatilization is not

120

Process Chemistry of Coal Utilization

implemented in discrete form across the CSTR-series. Rather, it is used in a standalone devolatilization simulation with FLASHCHAIN® to determine the timeresolved yields of all the important volatile species. The product yield histories are then subdivided into increments for the mean residence time of each CSTR in the series for the near-injector region under consideration. In other words, the fuel fed into a near-injector CSTR-series is a mixture of char and volatiles, where the volatiles are added in increments assigned for the mean transit times of individual reactors from the stand-alone devolatilization simulation; however, the total char flow is injected into the first CSTR. Except for the heat transfer zones at the outlets of fuel injectors, the particle thermal histories from the CFD simulation are not incorporated into the calculations with detailed chemistry, because that would compromise the benefits of the comprehensive mechanism for char oxidation in the ChemNet calculations. During char oxidation, the instantaneous particle temperature represents the interplay among numerous heat transfer mechanisms, including thermal inertia, convection, radiation, and heat release due to char oxidation. Accordingly, the temperature and burnout histories of particles are assigned from coupled balances on particle mass, size, and enthalpy within CBK/E. The radiation flux in the enthalpy balance contains the effective ambient temperature and the convective heat flux contains the mean gas temperature, both of which are functions of the mean transit time throughout the region under consideration. The final operating conditions to be specified are the entrainment rates into all regions. When the region under consideration is an injector, the flowrates of fuel and air into the region are unambiguous because there is only a single point of entry. However, for mixing layers, relatively thin zones for char burnout, OFA injection elevations, and other regions in which two or more streams mix, all flow into the region must be quantified, including entrainment. These entrainment rates must also be cast as functions of the mean transit time across the region for consistency with the other operating conditions. For regions of simpler, axisymmetric shapes, the entrainment rates may be evaluated from the analytical definition for the turbulent flux across the boundary of the region. More generally, fluid particles are tracked through the surrounding flows over the entire surface of the region under consideration. The tracking directly indicates the flowrate entering the region, which is interpreted as the entrainment rate. The total entrainment flowrate is then distributed in time, based on the flowrates through particular locations on the regional boundary compiled in the particle tracking. The entrained fluid is instantaneously dispersed over the cross-section of the region, in the direction transverse to the nominal flow (time) coordinate, as imposed by the governing equations for CSTRs and PFRs. This approach bases the entrainment rates on the multidimensional gradients and turbulent transport rates in the CFD simulation yet remains compatible with the Lagrangian trajectory in the reactor network calculations. ChemNet CFD postprocessing is contrasted with conventional CFD postprocessing in Fig. 5.3. In conventional postprocessing, the primary CFD simulation determines fields of velocity, temperature, and concentration based on rudimentary reaction rate expressions, and these fields are analyzed with additional global reactions in a second calculation pass. The second calculations do not improve or otherwise modify the

Simulations with detailed chemistry

121

Fig. 5.3 Schematic of the information flow in conventional and ChemNet CFD postprocessing.

results from the primary CFD; they simply implement rudimentary global reaction schemes to describe pollutant formation. In ChemNet postprocessing, the primary CFD is the same, except that additional field variables, such as turbulent diffusivities, particle concentrations, and conserved scalars, are in the output. The second step in the analysis processes only the primary flow and temperature fields with the supplemental output to develop an equivalent reactor network for the reaction system. The network is equivalent in that it imposes the same RTDs, dynamic thermal histories for gas and the effective radiation source, and dynamic entrainment rates as the CFD flowfield. The third calculation pass implements chemical reaction mechanisms of essentially unbounded complexity across the network to simultaneously simulate the conversion of volatiles and char and pollutant formation. ChemNet’s main advantages are that fully validated elementary chemical reaction mechanisms are implemented without approximation to describe volatiles conversion, NOX production and abatement, and residual CO levels in the flue gas. In combination with comprehensive reaction mechanisms for devolatilization and char conversion, the analysis takes full advantage of the phenomenal advances in coal conversion chemistry over the past few decades. While the analysis supersedes the rudimentary chemistry in CFD, it completely omits the impact of turbulent fluctuations on the chemistry and all other aspects of the flowfield.

5.3.2 Implementation protocol To summarize the previous section, the development of an equivalent reactor network proceeds through the following sequence of steps: (1) The CFD flowfield is delineated into regions whose chemistry is distinctive. The actual basis for the delineation may be the local concentrations of combustibles, especially soot and fuel particles, or a temperature field that can specify a meaningful average thermal history, or by an abundance of oxidizer and no fuel, which essentially suppresses the chemistry. (2) The RTDs of each region are determined from the CFD simulations by fluid particle tracking. Each RTD is then assigned a sequence of reactor elements, usually by fitting the analytical RTD for a CSTR-series to specify the number of CSTRs for the RTD under consideration.

122

Process Chemistry of Coal Utilization

(3) An average gas temperature history for each region is evaluated from the CFD gas temperature field by fluid particle tracking. The average history is then implemented in discrete form across the CSTR-series under consideration. (4) An effective ambient (wall) temperature for radiation transfer is evaluated as an average over the surrounding sections around the region under consideration. It is also implemented in discrete form across the CSTR-series. (5) If the region connects to a fuel injector, an average particle temperature history is assigned as the average of the thermal histories over all particle trajectories over the cross-section of the fuel suspension, so that the fuel’s devolatilization behavior can be evaluated. The predicted volatiles yields are fed as discrete injections of molecular species into all CSTRs whose mean transit times include portions of the predicted devolatilization period. All char is fed into the first reactor in the network. (6) Entrainment rates into all regions are evaluated as functions of the nominal time coordinate through the region under consideration. These rates are specified from the total mass flux across the boundaries of the region, for simple shapes, or from fluid particle tracking from the surroundings into the region, in the more general situation.

The analyses associated with each step are developed in the next sections.

5.3.2.1 Delineating regions Subdivision of the CFD flowfield into regions with distinctive chemistry is the first step in ChemNet postprocessing. Regions near fuel injectors are identified from the extents of mixing between the fuel suspension and any secondary air streams (as primary fuel jets are premixed with primary air). To quantitatively characterize mixing near fuel injectors, a conserved scalar tracks the mass fraction of all combustible material (C, H, O, N, S) in both the particle and gas phases, normalized by the inlet value, which is defined as: ψ ðx, y, zÞ ¼

mchons ðx, y, zÞ mchons,0

(5.1a)

where ψ is the local combustibles mass fraction; mchons is the mass fraction of combustibles at any position; and mchons,0 is the combustibles mass fraction at the inlet plane of the injector. The mass fraction of combustibles is calculated from mass fractions of volatiles, CO2, CO, H2O, NO, SO2, and the daf mass concentration of particles, as follows:
 12 12 14 2 32 ρdaf + ρ mvol + mCO2 + mCO + mNO + mH2 O + mSO2 44 28 30 18 64 mchons ¼ ρdaf + ρ (5.1b) where ρ is the local gas density (kg/m3); mvol, mCO2, mCO, mNO, mH2O, mSO2 are mass fractions of volatiles, CO2, CO, NO, H2O and SO2, respectively. The combustibles

Simulations with detailed chemistry

123

mass fraction includes the combustible elements, regardless of phase and regardless of whether they appear in reactants, intermediates, or products. ρdaf is the daf concentration of particles (kg/m3) given by np X

ρdaf ¼

i¼1

ðmi  mi0 fash Þ (5.1c)

Vcell

where np is the total number of particles in a computational cell; mi is the remaining mass of particle i; mi0 is the initial mass of particle i; fash is the ash fraction in the original coal particles, and Vcell is the control volume. This variable is not normally evaluated in a CFD simulation and therefore requires a UDF. As the combustibles mass fraction is a conserved scalar, its local value is determined by the convective and diffusive transport mechanisms in the CFD simulation. Sources and sinks, such as chemical reactions, do not affect the value. As such, the magnitude of the combustibles mass fraction diminishes in proportion to the entrainment of surrounding fluid into the primary fuel streams, and to the dispersion of combustibles away from the primary fuel streams. As an illustration, the regions delineated from a CFD simulation of a 1.7 MWth pilot-scale coal flame operated to simulate the emissions from full-scale T-fired furnaces appear in Fig. 5.4. The flame is attached to a weakly swirled, single-register burner. The near-burner flame structure consists of a fuel-rich flame core surrounded by a mixing layer between the core and the secondary air stream. Whereas the particles in this flame remain very close to the furnace axis, fluid from the core enters the mixing layer where it contacts secondary air and burns. In this flame, the core was defined as the locus of points where the combustibles mass fraction exceeds 0.4, and the extent of the mixing layer was based on a combustibles mass fraction of 0.3, which is the well-mixed value for the primary and secondary streams in this burner. Cores are generally smaller than their primary fuel jets, because the combustibles are concentrated in the center of the jet, according to the normal distribution of species concentrations in cross-section with entrainment of a surrounding fluid.

ERZ

Core

Mixing Layer

Fig. 5.4 Regions in a 1.7 MWth pilot-scale furnace.

OFA Zone

Burnout Zone

124

Process Chemistry of Coal Utilization

Cores often begin at some residence time after injection, rather than at the injection plane, especially when injectors deliver the fuel suspension rather than swirled burners. The intervening region between the injection plane and the beginning of the core is called the attachment region. Even though the combustibles mass fraction in the attachment region satisfies the threshold on the combustibles mass fraction, this region can be omitted from the chemistry calculations based on its cool particle temperatures. Attachment regions can be tagged with combustibles mass fractions exceeding 0.4 and mean gas temperatures that remain below 450°C. All attachment regions are nonreactive, by definition. Secondary air contacts the fluid from the core in a mixing layer that remains thin over most of the core length, but then fans out over the entire cross-section beyond the tip of the core. Almost all secondary air mixes with the core flow downstream of the core tip. A relatively thin ERZ fills the upstream furnace corners, pulling products from the mixing layer into the upstream secondary air stream. However, this region does not contain appreciable levels of gaseous fuel components or char particles, so it could be omitted from the ChemNet calculations. Downstream of the mixing layer, OFA air is injected through four off-radius jets. Some eddies alter the gas flow streamlines near the injection ports, but particle trajectories are hardly affected by the OFA flows. The OFA injection zones were delineated with O2 mass fractions above a threshold value. As they do not support any chemistry, they are also omitted from the ChemNet calculations. Downstream of the OFA regions, the flow is essentially a PFR with minimal gradients in temperature and particle loading. This so-called burnout zone is delineated as the entirety of the flow downstream of the OFA regions. It sustains the later stages of char oxidation during 1.5 s of transit time before the flue gas passes through a convective section and gas cleaning system.

5.3.2.2 Residence time distributions Once a boundary for a region has been delineated, the residence time of a single fluid particle is easily evaluated with fluid particle tracking, based on the difference between the initial time that a fluid enters the region and the elapsed time to its departure. Transit times are recorded for all fluid particle tracks in the analysis. A statistical analysis compiles an RTD from the transit times for individual trajectories. The CFD-based RTDs then specify the number of CSTRs in a series that will represent the region under consideration in the equivalent reactor network. The section of the network for a specific region is called an “equivalent reactor assembly.” CSTR-series are proposed first for all reactor assemblies but, occasionally, more complicated configurations are necessary. CSTR-series are emphasized because the CSTR-number in the series is easily determined from a least-squares fit of the following analytical expression to the CFD-based RTD: RTDðtÞ ¼


N1
 1 t t exp  ðN  1Þ!ti ti ti

(5.2)

Simulations with detailed chemistry

125

where RTD(t) is the exit age distribution of fluid in the region as a function of time, t; N is the number of CSTRs in the series; and ti is the mean residence time of an individual CSTR. All reactors in the series have the same properties, except volume. The assignment of N in the least-squares fit to the CFD-based RTD is particularly efficient because only integer values are acceptable. Cases that have N greater than about 25 during the analysis are aborted to avoid overflows and interpreted as PFRs although, in practice, the CSTR-number for which the predictions become insensitive to the CSTR-number is determined by the governing chemical reaction mechanisms for the particular region under consideration. RTDs for the core, mixing layer, and burnout zone determined with fluid element tracking throughout the CFD flow field appear in Fig. 5.5, along with their analytical representations. The RTD for the core was deconvoluted into one component for 13 CSTRs-in-series and another for plug flow with respective mean residence times of 138 and 192 ms. The plug flow component represents the near-axial fluid motion under the influence of particle drag, and the CSTR-component represents flow with significant radial velocities. The RTD for the mixing layer was matched with a series of 19 CSTRs, and that for the OFA zone (not shown) was represented by 6 CSTRs-inseries. The burnout zone is essentially in plug flow. Since the CFD-based RTDs determine the mean residence time for each region and the analytical fits to each RTD determine a CSTR-number, the mean transit time for each CSTR is evaluated as their ratio.

Fig. 5.5 RTDs (bars) from the CFD simulation and (curves) assigned from Eq. (5.2) for the core, mixing layer, and burnout zone of the pilot-scale flame with 15% OFA and 3.5% O2.

126

Process Chemistry of Coal Utilization

5.3.2.3 Operating conditions The mean gas temperature histories can be simultaneously assigned with the RTDs. The time scale for the temperature history is specified with 50 equal increments of the longest populated residence time in an RTD. The longest populated time is evaluated as the time in the RTD which was longer than the residence times of 95% of the individual fluid particle transit times. Then, the temperatures along the trajectories of individual fluid particles are recorded in the same time increments. This operation puts the individual temperature histories on a consistent time scale for averaging. The gas temperatures for all fluid particles in the region are averaged at each time increment, according to: X Nf  T̄ g,i ¼

j¼1

u0, j Tg,i, j



XNf

(5.3)

u j¼1 0, j

where subscript i represents the ith time increment in the time scale and j represents the jth fluid trajectory; Nf is the total number of fluid tracks through the region; and u0, j is the fluid velocity at the injector, which is used to mass-weight the average, since the gas density at the injection plane is uniform because the temperatures are uniform. The CFD-based and discrete rendition of the gas temperature history for the core of the pilot-scale flame appears in Fig. 5.6. This particular determination was based on a series of 16 CSTRs with a total transit time of 163 ms. Each temperature in the CFD-based history was determined as the average temperature across the core, transverse to the flow direction. To specify the discrete version, this thermal history was subsequently averaged over each increment in residence time for the CSTR-series that represents the core. As the incremental residence time is only 10 ms, the discrete rendition depicts the CFD-based thermal history without undue uncertainty. Thermal 1600 1400

Particle Temperature, °C

Gas Temperature, °C

1500 hv bituminous, Core 15% OFA, 3.5% O2 1250

hv bituminous 15% OFA, 3% O2

1200

1000

1000

750 500 250 0 0.00

800 600 400 200

0.05

0.10 0.15 Residence Time, s

0.20

0 0.000

0.025

0.050 0.075 Residence Time, s

0.100

Fig. 5.6 (Left) Gas temperature history (histogram) assigned to the CSTR series for the core compared to (solid curve) the continuous history assigned from the CFD simulation. (Right) Mean particle temperature histories assigned for devolatilization simulations in the core (dashed curve) from CFD and (solid curve) adjusted for FLASHCHAIN® simulations. The histogram shows the time increments for volatiles injection into the CSTR-series for the flame core.

Simulations with detailed chemistry

127

histories for radiation temperature are assigned the same way, and both histories are assigned for all regions in a furnace or gasifier with the same method. In addition to the ambient thermal history, an average thermal history of particles through the core is needed for the devolatilization simulation. The history used in the simulation is compared to the CFD-based history in Fig. 5.6. There is a nonphysical lag in the CFD-based history through 10 ms which was deliberately omitted in the assigned history. Even so, the heating rates for both cases are too similar to affect the predicted devolatilization behavior. The discrepancy for times longer than 60 ms is also inconsequential because devolatilization is complete by this time. A histogram of incremental transit times for the associated CSTR-series is superimposed on the thermal histories. This does not imply that the devolatilization simulation is based on a discrete rendition of the particle temperature history. Rather, the cumulative volatiles yields and compositions are apportioned to each increment in residence time to specify the increments of volatiles injected into individual reactors in the CSTR-series. The entrainment of oxidizing streams into fuel-rich streams is crucial for accurate predictions of furnace combustion efficiencies and flue gas compositions. In furnaces where the regions have simple shapes and regular boundaries (such as the pilot-scale coal flame), entrainment rates can be directly specified from the entrainment flux across the boundary of a region. This flux, QO2 in kg/m2 s of O2, is defined as follows:     * * ! QO2 ðsÞ ¼ ρDeff , O2 n  rmO2 + ρmO2 v  n

(5.4a) *

where mO2 is O2 mass fraction; Deff, O2 is the effective (turbulent) diffusivity of O2; n is * the unit vector normal to the boundary; and v is the velocity vector. The two terms on the right-hand side of Eq. (5.4a) represent diffusion and convection of O2, respectively. The entrainment rate m_ O2 in kg/s is evaluated by integrating the entrainment flux over the entire boundary, according to ðð m_ O2 ¼

QO2 ðsÞds

(5.4b)

s

where s is the boundary of the region. These expressions can be evaluated with flow visualization software, then the assigned entrainment rates can be transformed onto the mean residence time coordinate used for the temperature histories. In situations where the region boundaries are too convoluted for this approach, entrainment rates are evaluated with fluid particle tracking. Fluid particle tracks from all air injection locations are generated from the velocity field in the CFD simulation. Thousands of tracks are usually required for each injector. Entrainment of these air streams into furnace regions with combustibles is analyzed to record where the fluid track crosses the boundary into a subject region, if it does. An overall entrainment rate is evaluated as a mass-weighted sum of all tracks that cross a boundary. The overall entrainment is distributed along the regional boundary according to the recorded locations of the penetrations from the fluid particle tracking. These locations are expressed

128

Process Chemistry of Coal Utilization

with the major coordinate axis that aligns best with the dominant flow direction of the region. For example, for the core regions from fuel injectors, the suspension axis is the dominant flow direction. This method works well for regions affected by convective processes. However, there are also regions in commercial furnaces where reactants are entrained by diffusive processes, such as the entrainment of O2 from the flow along furnace walls above the SOFA injectors into the central char burnout region in commercial furnaces. For such situations, the convective and diffusive velocity vectors for O2 are formulated from the CFD simulation based on a rearrangement of Eq. (5.4a). The turbulent diffusivity in the definition is taken directly from the CFD simulation. Then the entrainment rate of O2 is evaluated as the component of this O2 flux in the transverse direction away from the furnace wall. The spatial entrainment profiles must be converted into time-histories for compatibility with all the other operating conditions and the sequencing in the detailed chemistry calculations. The time coordinate in this history must be the same as in the temperature histories, so the positions on a region’s surface along the dominant flow direction that was used to assign transit times in the gas temperature histories are used for entrainment without modification. For the pilot-scale flame, entrainments of secondary air into the mixing layer and OFA into the process flow were first evaluated directly from the CFD simulation with the method based on Eqs. (5.4a), (5.4b). The entrained fraction of secondary air into the mixing layer appears in Fig. 5.7. The entrainment rate decays exponentially across the mixing layer, with substantial irregularities. The surge during the first 100 ms is due to the intersection of the secondary air stream with the expanded primary coal suspension, which expands due to the fast heating rate of the suspension. Ultimately, the entrainment fraction reaches an asymptotic value of 0.84, which is the portion of the secondary air not immediately entrained into the primary stream near the fuel injector outlet. The remainder was aspirated directly into the primary jet at the burner outlet plane by a pressure imbalance.

0.8

15 Entrainment Volume

Fraction Entrained 2° Air

1.0

0.6

0.4

Jet in Co-Flow 10

Jet in Cross-Flow

5

0.2

0.0 0.0

0 0.1

0.2

0.3 0.4 Time in ML, s

0.5

0.6

1

2

3 4 5 6 Distance From Injection

7

8

Fig. 5.7 (Left) Fraction of entrained secondary (2°) air into the mixing layer from the CFD simulation for the pilot-scale flame with 15% OFA and 3.5% O2; and (right) entrainment volumes for two limiting mixing models versus distance from the point of injection.

Simulations with detailed chemistry

129

Air entrainment is incorporated into the ChemNet simulations with either of two simple mixing models, whose behavior is illustrated in Fig. 5.7, right panel. The curves in this figure depict the volumetric air entrainment rate at each point downstream of a common injection point. For jets in co-flow, the entrainment volume grows with distance from the injection point, according to the following functional form: FE ¼ exp ðβM ðt  τM ÞÞ  1 for t  τM

(5.5a)

where FE is the fraction of the air stream that has been entrained into the process flow to time t; τM is a time lag equal to the nominal residence time to the injection point; and βM is an empirical mixing constant. According to Eq. 5.5a, the entrainment rate grows with distance from the injection point until the entire secondary stream has been entrained. For jets in cross-flow, the entrainment volume diminishes with distance from the injection point, according to FE ¼ 1  exp ðβM ðt  τM ÞÞ for t  τM

(5.5b)

This entrainment rate diminishes with distance from the injection point until the entire secondary stream has been entrained. Once the parameters in one of these mixing models are assigned by fitting an entrainment history, the continuous profile of entrainment fraction versus time from the injection point is resolved into discrete additions to each CSTR in a series. As the entrainment of secondary air into the primary air stream at the burner outlet is very rapid, no finite-rate mixing models were applied to this portion of the secondary air stream. That entrainment was simply added to the initial primary air stream. To summarize, the mean transit times and the CSTR-number for each region are assigned from the CFD-based RTDs. Their ratio specifies the nominal transit time for each CSTR. Temperatures are assigned to each isothermal CSTR from the CFD-based gas thermal histories. The effective radiation temperature in the energy balance for burning char is specified in the same way from the wall temperature history. But particle temperature histories do not need to be allocated to individual CSTRs, because a stand-alone devolatilization simulation can use the continuous, CFD-based mean particle temperature history. However, the release of volatiles and their compositions are rendered into discrete increments for each CSTR immediately downstream of the fuel injector. Rapid, near-burner entrainment of secondary air simply supplements the primary air stream. But one of two finite-rate mixing models is fit to the CFD-based entrainment histories and implemented in the ChemNet simulations for all downstream regions.

5.3.2.4 Required variables for ChemNet postprocessing The CFD simulation used to specify an equivalent reactor network in a ChemNet analysis is conventional in the engineering sense. It should incorporate the recommendations to improve the chemistry submodel in Chapter 4 but, otherwise, requires no

130

Process Chemistry of Coal Utilization

specialized physicochemical mechanisms. However, a ChemNet analysis does require specialized output from the CFD simulation, as described in this section. Among the field variables for gases, the postprocessing uses gas density and temperature; the mass fractions of all species; effective turbulent mass diffusivities; three components of gas velocity; effective radiation temperature; and the mass concentration of particles, in kg/m3. Only the mass concentration of particles requires a UDF to evaluate it from other field variables. Particle trajectories are retrieved under a “Particle Track” output option and provide three coordinates that locate each particle, plus a residence time, temperature, mass, and size. The information from the field variables for gases and the particle trajectories is sufficient to evaluate the conserved scalar variables in Eqs. (5.1a)–(5.1c). The gross operating conditions also need to be specified, including mass flowrates of all inert and oxidizing streams, and all fuel and sorbent fuel streams. Gasifiers may also have steam injection rates. Proximate and ultimate analyses and PSDs are required for all fuels and fuel blend components. From a practical standpoint, a commercial flow visualization package is essential to interrogate and analyze the very large populations of physical and fluid particles from the CFD. Moreover, the analysis is labor-intensive, so the benefits from such a highly accurate flow characterization must be seen clearly in advance. Conversely, many important questions can be answered with networks as simple as plug flow reactors that depict only the grossest features of actual flow patterns. Before a practitioner commits to a thorough analysis to specify an equivalent network with the highest fidelity to the CFD simulation, it is often worthwhile to assess the chemistry with a radically simplified network, just to make sure that the species concentrations of interest will actually be clarified by a ChemNet analysis with full chemistry.

5.4

Chemical reaction mechanisms

Once the equivalent reactor network has been specified, the chemistry in each reactor in the network is sequentially evaluated from the species balances based on elementary reactions for the gas phase and soot. This section briefly reviews the reaction mechanisms that were implemented in the case studies in Chapters 6 and 7. All mechanisms except the surface reaction mechanism for soot have been independently validated over most of the full domain for coal processing, as described in the first volume of this series. Consequently, only the soot chemistry submodel is described in detail. Of course, alternative mechanisms for any stage of the chemistry would be suitable for ChemNet postprocessing, provided they were validated for the domain of operating conditions in the application.

5.4.1 Fully validated mechanisms A multitude of fuel species—CO, H2, CH4, C2H2, HCN, H2S, oils, tar, soot, and char—compete for the available O2 in a pc flame. This competition determines local heat release rates, which govern flame stability, combustion efficiency, UBC, and the

Simulations with detailed chemistry

131

local oxidizing potential of the gas phase which, in turn, governs N-species conversion. The crucial outcome of this competition cannot be forecast from the burning rates of the individual fuels determined in isolation. Instead, realistic chemical kinetics for each distinctive conversion process must be incorporated into a comprehensive analysis. A ChemNet analysis incorporates the most comprehensive chemical reaction mechanisms available and imposes no a priori assumptions regarding the apportioning of O2 in combustion systems, or the gasification agents and inhibitors in gasifiers. The mechanisms for coal conversion are the ones used to specify the parameters in CFD chemistry submodels in Chapter 4: FLASHCHAIN®-based mechanisms describe primary devolatilization and the decomposition or hydrogenation of primary tar; CBK/E describes char oxidation; NSC kinetics describe soot oxidation; and CBK/G describes the gasification of both char and soot. The FLASHCHAIN®-based mechanisms distinguish primary devolatilization, which relates fuel properties to the composition of volatiles, from secondary volatiles pyrolysis, which generates the volatiles that actually burn in pc flames. FLASHCHAIN® determines the complete distribution of primary products from any coal and also predicts the yield and elemental composition of char. When combined with a swelling factor correlation and a correlation for the initial carbon density in char, it specifies all the necessary char properties for a char conversion simulation. The complete distribution of volatiles, including gaseous fuels and soot, and all properties of char and soot are completely determined from only the coal’s proximate and ultimate analyses. The reaction mechanism for chemistry in the gas phase must describe the ignition, reforming, and oxidation of all secondary volatiles pyrolysis products, as well as the conversion of all N-species across the full range of SR values in pc flames, and under the deeply reducing conditions in pyrolyzers and gasifiers. The homogeneous mechanism of Glarborg et al. (1998) was used in nearly all case studies for pc furnaces in Chapter 6, and for some gasifier case studies in Chapter 7. It contains 444 elementary reactions among 66 species, including all relevant radicals and N-species. It is implemented in the simulations without any approximations whatsoever. Most simulations on pyrolysis and gasification in Chapter 7 implement an elementary mechanism developed for diesel combustion applications, to contend with the elevated pressures. This alternative mechanism has 566 reactions among 154 species, including all GHCs up to benzene, toluene, and phenol. It combines submechanisms for C3 GHCs (Curran et al., 1998; Seiser et al., 2000) with one for single-ring aromatics (Pitz et al., 2003), and was validated for pressures from 0.1 to 4.2 MPa; temperatures from 280 to 1430°C, and equivalence ratios from 0.3 to 1.5. However, calculations have demonstrated that Glarborg et al.’s homogeneous mechanism gives essentially the same product compositions as the alternate mechanism and, consequently, all results in this book would be essentially the same with either homogeneous mechanism. With either mechanism, all rate parameters were assigned independently, so there are no adjustable parameters in the mechanisms for gas-phase chemistry. Soot plays several important roles. As it burns, it directly competes for the available O2 and also consumes O-atoms and OH that would otherwise sustain homogeneous chemistry. Soot also promotes recombinations of H-atoms and OH that could also sustain homogeneous chemistry. And soot reduces NO directly into N2. The soot

132

Process Chemistry of Coal Utilization

chemistry submodel presented in the succeeding section depicts all these effects in the form of a collection of elementary reactions that can be coupled to the homogeneous reaction mechanism. Char burning rates are determined by surface chemistry, intraparticle pore transport, thermal annealing, ash encapsulation (of low-rank chars), and a transition to chemical kinetic control. With coal chars, CBK/E accurately predicts combustion histories for char across the full domain of temperature, pressure, gas composition, size, and coal quality in suspension firing. This performance is based on a three-step surface reaction mechanism plus an energy distribution for thermal annealing; plus transport mediation by both film and intraparticle diffusion; plus an ash encapsulation mechanism that inhibits reactant penetration during the latter stages of char burnout. The mathematics behind these mechanisms and the energy balance to evaluate a char particle’s thermal history throughout burnout have been reported (Niksa et al., 2003). However, it is not yet possible to specify the initial char reactivity for a particular char sample within useful tolerances from the standard coal properties. One-point calibration is required to specify this value with measured LOI or some other suitable index on combustion efficiency. The submodel for char-N conversion is subject to a similar calibration requirement (with NO emissions), compounded by its simplistic mechanistic premise: that a fixed fraction of char-N is converted into NO at the overall char burning rate throughout all stages of char oxidation. Char gasification rates are determined by complex surface chemistry, plus the evolution of a char’s pore structure, plus the thermal annealing and ash encapsulation that affects char burning rates. With coal chars, CBK/G accurately predicts gasification histories across the full domain of temperature, pressure, gas composition, size, and coal quality (Liu and Niksa, 2004). This performance is based on a five-step surface reaction mechanism plus an energy distribution for thermal annealing; plus transport mediation by both film and intraparticle diffusion; plus morphology changes from the Random Pore Model (RPM); plus an ash encapsulation mechanism that inhibits reactant penetration during the latter stages of char conversion. The surface mechanism describes gasification by steam and CO2 with inhibition by H2 and CO and hydrogasification. CBK/G covers variations in gas temperature, the levels of gasification agents and inhibitors, and char particle size within useful quantitative tolerances. However, it is not yet possible to specify the initial char gasification reactivity within useful tolerances from the standard coal properties. One-point calibration is required to specify this value with measured extents of char conversion. Soot gasification is analyzed with the CBK/G mechanisms for surface chemistry and annealing, but without pore transport, pore evolution, and ash encapsulation. Surface kinetics for soot are comparable to those for coals of the highest rank, because the carbon contents are comparable. These chemical reaction rates are much slower than those for most chars, because soot contains little, if any, alkali and alkaline earth metals that catalyze gasification; nonetheless, soots’ gasification rates are comparable to chars’ due to the much smaller particle sizes. The mechanisms are collected in Table 5.1. To summarize their status, the mechanisms for devolatilization, tar decomposition, homogeneous chemistry, char conversion, and soot oxidation are complete, whereas those for soot/radical chemistry and

Simulations with detailed chemistry

133

Table 5.1 Features of the reaction mechanisms.

Devolatilization Fuel type Yields Composition Secondary volatiles pyrolysis

Any coal or petroleum coke and all forms of biomass (woods, grasses, and agricultural residues). Sample-specific predictions from FLASHCHAIN® based on only the proximate and ultimate analyses. Gases resolved as molecular species. C/H/O/N/S compositions of char and tar. Both instantaneous and finite-rate tar decomposition with ultimate products of oils, PAH, and/or soot. Both scenarios have a simultaneous release of CO, H2, and HCN and destruction of all hydrocarbons except CH4 and C2H2. C/H/N composition of soot.

Char properties Composition Size changes Bulk density Carbon density

C/H/O/N/S composition from FLASHCHAIN®. Swelling-factor correlation assigns the char PSD from the fuel PSD. Determined from volatiles yield and the swelling factor correlation. Correlated with the C-content of the parent coal.

Homogeneous chemistry Oxidation Reforming

444 Elementary reactions among 66 species, including all products of secondary volatiles pyrolysis and N-species. 566 Elementary reactions among 154 species, including all GHCs up to benzene, toluene, and phenol.

Soot chemistry Oxidation Recombinations NO reduction

The four-step elementary reaction mechanism for oxidation by O2 matches NSC kinetics. Simultaneous oxidation by O and OH. Recombinations of H into H2 and OH into H2O. Single quasiglobal process.

Char conversion Oxidation

Char-N conversion

Gasification

Burnout from CBK/E with transitions through burning regimes, thermal annealing, and ash encapsulation. NSC kinetics for soot oxidation. Initial oxidation reactivity for a coal char should be calibrated with data. The fixed portion of char-N converted into NO at the nominal burning rate should be calibrated with baseline NOX emissions. Conversion based on CBK/G for steam and CO2 with inhibition by H2 and CO plus hydrogasification, including thermal annealing and ash encapsulation. Initial gasification reactivity should be calibrated with data.

134

Process Chemistry of Coal Utilization

char-N conversion will probably be subject to important revisions in the future. Neither of these latter two situations introduces significant uncertainties into ChemNet simulations of furnaces and gasifiers. As these reaction mechanisms have already been independently validated across an enormous domain of conditions, what matters most is the degree to which all model parameters can be specified from the available information on the furnace operating conditions. The initial char reactivities for oxidation and gasification and the fraction of char-N converted to NO during burnout can only be specified from calibration procedures, whereby these parameters are adjusted to match the predicted LOI and NOX emissions to reported values for a single set of operating conditions. Then the same values should be imposed for all other operating conditions. Except for these three parameters, all other model parameters can be assigned from the fuel’s proximate and ultimate analyses within useful quantitative tolerances, or directly adopted from literature. It is no small irony that realistic, comprehensive reaction mechanisms have many fewer adjustable parameters than rudimentary global reaction schemes. Collectively, the global rates in a conventional CFD chemistry submodel contain a few dozen kinetic parameters that must be adjusted for every coal. In contrast, the suite of comprehensive mechanisms in this section contains just three: initial char reactivities for oxidation and gasification plus the fraction of char-N converted to NO. Consequently, the amount of calibration data required to depict the distinctive behavior of a particular coal sample is far smaller with the comprehensive mechanisms. And the only sample-specific information they require is proximate and ultimate analyses for every coal sample.

5.4.2 Surface reaction mechanism for soot chemistry Surface reaction mechanisms on soot were introduced into the analysis of pc coal flames by Pedersen et al. (1998b). They combined a multistep surface oxidation/ recombination mechanism on soot with a simplified mechanism for volatiles combustion developed from a reduced version of a complete reburning chemistry mechanism. This homogeneous mechanism omits all C1/C2 chemistry and regards volatiles as H2/CO mixtures, but otherwise retains realistic submechanisms for fuel oxidation and N-species conversion. Qualitatively, the predicted behavior is the same as with the full reburning mechanism. The submechanism for soot oxidation accounts for radical recombinations, oxidation by O and OH, and NO reduction on soot. The calculations with and without the soot chemistry submechanism are especially informative, as seen in Fig. 5.8. Reactions on soot shift the SR transition at which NO production begins toward leaner mixtures. The effect occurs for two reasons: (1) Soot oxidation consumes O2 which would otherwise participate in the homogeneous reaction mechanisms; and (2) H-atoms and OH-radicals recombine on soot, thereby slowing the burning rates in the gas phase like any other termination reaction. In addition, NO may be reduced on the carbon in soot. Since then, homogeneous reaction mechanisms in conjunction with comprehensive coal conversion mechanisms but without soot chemistry have been able to accurately depict how the N-speciation shifts toward NO production for progressively greater

Simulations with detailed chemistry

135 2500

H2 OH

1500

4 1000 2

O

Soot

500

6 Mole Fraction, ppm Mole Fraction, %

Mole Fraction, %

No Soot Chemistry 2000

6

2500

8

With Soot Chemistry

CO

OH

H2

2000

1500 CO

4

O 2

H

Soot

1000

Mole Fraction, ppm

8

500

H 0

0

No Soot Chemistry 2500 TFN

2000

2000 1500

1500 TFN Gas Only

1000

1000 HCN 500 0 0.0

TFN Gas Only NH3 0.2

NO 0.4

0.6

HCN

NO 0.8

1.0

0.0

0.2

0.4

Local SR

0.6

0.8

Mole Fraction, ppm

Mole Fraction, ppm

2500

0

0

With Soot Chemistry

TFN

500 0 1.0

Local SR

Fig. 5.8 Model predictions for (left) the complete soot surface/homogeneous mechanism and (right) without soot/radical reactions. Reproduced from Pedersen LS, Glarborg P, Dam-Johansen K. A reduced reaction scheme for volatile nitrogen conversion in coal combustion. Combust Sci Technol 1998b;131:193–223 with permission from Taylor and Francis.

SR. Moreover, the predicted N-speciation was within measurement uncertainties of reported speciations for diverse coal types (Niksa, 2019a). Consequently, the role of surface chemistry on soot is ambiguous because test measurements to discriminate among alternative modeling approaches on this topic are not yet available. The author’s soot chemistry submechanism in the next section is used in some, but not all, of the case studies in Chapter 6. Pending stringent validation work, it is presented as an option for practitioners to consider for their simulations.

5.4.2.1 Soot chemistry submechanism The rates of soot oxidation in the presence of O2 for temperatures to 3000 K and pressures to 30 atm are well-described by NSC kinetics, which are ωNSC ¼

kA pO2 χ + kB pO2 ð1  χ Þ where χ ¼ 1 + kZ pO2

1 1+

kT k B pO 2

(5.6)

The soot oxidation rate, ωNSC, has units of g/cm2 s. It is determined by four rate constants of Arrhenius form, kA, kB, kT, and kZ, and the local free-stream partial pressure of

136

Process Chemistry of Coal Utilization

O2, pO2. This expression is semiempirical, in that the authors interpreted the rate constants in terms of elementary reaction processes but did not propose a surface reaction mechanism that determines the oxidation rate in Eq. (5.6). NSC-kinetics are robust enough for applications in coal conversion simulations. But they are difficult to implement with elementary reaction mechanisms because they do not conform to the format for rate expressions in popular chemical kinetics packages that solve the ordinary differential equations (ODEs) for the chemistry in CSTRs and PFRs. In lieu of a customized ODE solver to handle both the chemistry-derived ODEs and the NSC reaction rate expression, the NSC rate expression will be expressed as a multistep sequence of elementary surface reactions, as follows. The elementary soot oxidation mechanism is developed in terms of three distinct surface sites, plus bulk carbon, SB, which replenishes surface sites: (1) SA denotes a reactive free site; SD denotes a nearly unreactive free site; and SO denotes a site with adsorbed O. There are four elementary reactions: kA

SA + O2 ! OðgÞ + SO kB

SB + SD + O2 ! COðgÞ + OðgÞ + SA kP

(5.R1) (5.R2)

SO + SB ! COðgÞ + SA

(5.R3)

SA ! SkDT

(5.R4)

Reactions (5.R1) and (5.R2) describe dissociative- and oxidative-chemisorption of O2 on the carbon sites. Oxidative chemisorption on an unreactive site directly produces CO whereas oxidative-chemisorption on a reactive site feeds into the desorption in Reaction (5.R3). Reactive sites are converted into nearly unreactive sites in Reaction (5.R4). If the fractional surface coverages, χ i, of all sites are in a steady state, then the total conversion is given by dCB ¼ kB PO2 ð1  χ A  χ O Þ + kP χ O dt ¼ kB PO2 ð1  χ A  χ O Þ + kA PO2 χ A

(5.7a)

where the second form follows from the steady-state assumption for χ O. The surface coverages are defined as kA PO ð1  χ O Þ kP 2 χA ¼ and χ O ¼ kT kA kT 1+ 1 + PO2 + kB PO2 kP kB PO2

(5.7b)

Simulations with detailed chemistry

137 –1

–1

–2 –3

–3 5 % O2

–4

–4

1 % O2

–5

–5 –6

0.1 % O2

–6 –7

log10RSOOT, g/cm2-s

25 % O2

–2 log10RSOOT, g/cm2-s

15 % O2

–7

0.01 % O2

–8

–8 3

4

5

6

7

8

3

4

5

104/T

6

7

8

104/T

Fig. 5.9 Soot oxidation rates from (solid curves) the elementary soot oxidation mechanism for O2 compared with (dashed curves) rates from NSC-kinetics that were assigned from measured values from shock tube tests (Park and Appleton, 1973).

If KAPO2/kP ≫ kT/kBPO2 and χ O ≪ 1, then the total conversion rate for the elementary reaction mechanism reduces to the NSC-kinetics in Eq. (5.6), as follows: ωSOOT ¼

kA PO2 χ + kB PO2 ð1  χ A Þ where χ A ¼ 1 + k Z PO 2 A

1

1+

k kT and kZ ¼ A kP k B PO 2 (5.7c)

Since the overall oxidation rate for this elementary oxidation mechanism has the same form as the NSC rate expression in Eq. (5.6), it is able to reproduce NSC kinetics. The values of the rate constants to achieve this in practice were specified by fitting the soot oxidation rates reported in Park and Appleton (1973) for shock tube tests. As seen in Fig. 5.9, the agreement is nearly exact for conditions relevant to pc combustion. Rates based on NSC kinetics generally agree with the shock tube measurements within experimental uncertainty. The agreement for the proposed surface mechanism is also generally within experimental uncertainties for temperatures to 2000 K and O2 concentrations to 25%, which covers the relevant domain of operating conditions in pc flames. For hotter temperatures, the predicted soot oxidation rates are generally faster than NSC-based values, although they are actually in better agreement with the measurements for some of the O2 concentrations in this evaluation. Rate constants specified in this evaluation are compiled in Table 5.2. In addition to oxidation by O2, soot also burns with O and OH; in fact, OH is often the primary soot oxidizer under rich-flame conditions (Neoh et al., 1981). These reactions were analyzed by Pedersen et al. (1998b), according to the following elementary reactions: kO

SB + SA + OðgÞ ! COðgÞ + SA

(5.R5)

138

Process Chemistry of Coal Utilization

Table 5.2 Soot chemistry submechanism. Form

A-Factor

T-Exp.

Ea (cal/mol)

Arrh. Arrh. Arrh. Arrh. Stck. Stck. Stck. Stck. Stck.

3.00E + 11 4.00E + 03 3.00E + 08 3.10E + 08 0.23 0.23 0.28 0.28 0.28

1 1 0 0 0 0 0 0 0

29,000 25,180 35,000 105,000 0 0 0 0 0

Stck. Stck. Stck. Stck.

0.90 0.90 0.19 0.65

0 0 0 0

0 0 2310 0

Stck. Stck.

1.82 1.82

0 0

29,800 29,800

Oxidation SA + O2(g)¼>) O(g) + SO SB + SD + O2(g) ) CO(g) + O(g) + SA SB + SO ) SA + CO(g) SA ) SD SB + SA + O(g) ) CO(g) + SA SB + SD + O(g) ) CO(g) + SA SB + 2SA + OH(g) ) CO(g) + SA + SH SB + 2SD + OH(g) ) CO(g) + SA + SH SB + SA + SD + OH(g) ) CO(g) + SA + SH

Recombination SA + H(g) ) SH SD + H(g) ) SH SH + H(g) ) H2(g) SH + OH(g) ) SA + H2O(g)

NO reduction SB + SA + NO(g) ) CO(g) + 1/2N2(g) + SA SB + SD + NO(g) ) CO(g) + 1/2N2(g) + SA

kOH

SB + 2SA + OHðgÞ ! COðgÞ + SH + SA

(5.R6)

where SH is the site of an adsorbed H-atom. In addition to these oxidation reactions, the following radical recombinations are also included: kH

SA + HðgÞ ! SH

(5.R7)

kH2

SH + HðgÞ ! SA + H2 ðgÞ kH2 O

SH + OHðgÞ ! SA + H2 OðgÞ

(5.R8) (5.R9)

The final soot reaction is the reduction of NO, according to kNO

SB + SA + NOðgÞ ! COðgÞ + 1=2N2 + SA

(5.R10)

This reaction is obviously not an elementary step due to the fractional stoichiometric coefficient for N2.

Simulations with detailed chemistry

139

The complete soot chemistry submodel combines the elementary reaction mechanism for oxidation by O2 with the radical oxidation steps plus the recombination and NO reduction reactions. In this combination, reactive (SA) and nearly unreactive (SD) sites were assumed to be equivalent sites for recombinations and NO reduction. The complete soot chemistry submodel is collected in Table 5.2, including the values of all rate constants. Two forms of rate constants are implemented. The ones denoted as “Arrh.” have a conventional Arrhenius form: ATnexp( Ea/RT). The three parameters in the table are the A-factor, the powerlaw exponent, n, and the apparent activation energy, Ea. Rate constants denoted by “Stck.” are based on sticking coefficients, which multiply the collisional frequency. Most of the sticking coefficients are simple multiplicative constants, although three include apparent activation energies which appear in the Boltzmann factors in the collision frequencies.

5.4.2.2 Implementation of soot chemistry Many of the computer applications that manage surface reaction mechanisms are normally used for catalytic reaction systems, in which the surface properties are fixed throughout the simulation. Some include an etching simulation mode, in which the surface either reacts away or grows by deposition. However, the assumption that the surface area and all other surface properties undergo negligible changes is usually embedded in this operating mode. This assumption breaks down in coal flame simulations whenever the extent of soot oxidation is significant across a particular reactor in the equivalent network. It can be circumvented with the iterative calculation scheme in this section. The total surface area of the soot in a reactor, ASOOT, is first expressed in terms of a uniform specific soot surface area, aSOOT, according to +1 ASOOT ¼ yjP,SOOT ρTOTAL VREACTOR aSOOT

(5.8)

where yP,SOOT j + 1 is the mass fraction of soot in the product stream leaving reactor j; ρTOTAL is the total density of material within the reactor; and VREACTOR is the reactor volume. The specific surface area is evaluated as 10 m2/g based on reported values for soots extracted from rich flames (Roessler et al., 1981), and fixed throughout the reactor simulation. The reactor volume is also fixed. The total density is only affected by the soot mass fraction, because purely homogeneous chemistry would not affect it under constant temperature and pressure. Consequently, ASOOT from Eq. (5.8) can only be determined after yP,SOOT j + 1 has been specified. With CSTRs, the contents of the reactor have a uniform composition that equals the product stream composition. So yP,SOOT j + 1 is the primary iteration variable, and ASOOT is assigned from its current value. The initial trial for ASOOT is based on the inlet composition, then successive iterations refine the trial with the calculated value for yP,SOOT j + 1 , until a convergence tolerance is satisfied.

140

Process Chemistry of Coal Utilization

At each iteration, the amount of soot that has been oxidized into combustion products is determined from the atomic C/O-ratio of the product gas composition. The ratio in the product stream is related to that in the reactant stream as follows:
j + 1
j ΔXj F∞ C C S =12 ¼ + j S,O2 j O O yP,O FP =16

(5.9a)

where superscripts j and j + 1 denote inlet and outlet conditions, respectively; ΔXjS, O2 is the fractional soot burnout across the reactor; FS ∞ is the mass flowrate of soot based on the ultimate soot yield; FP j is the inlet mass flowrate of gases; and yj P,O is the mass fraction of oxygen in the inlet stream. Eq. (5.9a) can be rearranged into the following explicit definition for ΔXjS, O2: j ΔXS,O 2

"

j # j yP,O FjP =16 C j+1 C ¼  O O F∞ S =12

(5.9b)

In the calculation sequence, the (C/O)-ratios are evaluated from the calculated gas compositions, then inserted into Eq. (5.9b) to specify the extent of soot burnout. Then the burnout value is used to calculate the mass flowrate of soot leaving the reactor, which factors into the suspension loading of soot and the soot mass fraction within the reactor. The soot mass fraction enters into the updated estimate for the total area of soot within the reactor, according to Eq. (5.8). When the procedure has converged, the assigned total soot area is consistent with the conversion and product composition, including soot conversion effects on both the total area within the reactor and the addition of carbon products to the gas phase.

5.5

Species conservation equations

Once an equivalent reactor network has been specified for the subject furnace or gasifier, species conservation equations are solved to determine the gas compositions leaving each CSTR, and the extents of conversion for char and soot under reactive atmospheres. As the thermal history for devolatilization is specified directly from the CFD results, and as each CSTR has a uniform temperature evaluated in the postprocessing, the only enthalpy balance that needs to be solved is for reacting char particles. This section develops the necessary species conservation equations for progressively more complex situations for suspension firing, beginning with nonreactive carrier gas streams, then with oxidizing streams, then with sequential combinations of oxidizing and reducing atmospheres for gasification systems. The enthalpy balances for char conversion have been reported elsewhere (Niksa et al., 2003; Liu and Niksa, 2004).

Simulations with detailed chemistry

141

5.5.1 Equations for nonreactive carrier gas streams The vast majority of applications of suspension firing inject coal in gas streams that contain O2. Even in pyrolyzers, O2 is usually present to raise process temperatures via partial oxidation. This section describes suspension firing with strictly nonreactive carrier streams to introduce the methodology and nomenclature for more realistic reaction systems in succeeding sections. The nomenclature is summarized in Fig. 5.10. The total mass flowrates of the process stream into and out of the jth CSTR in the network are FP j and FP j + 1 , respectively. FC j and FC j + 1 are the corresponding mass flowrates of char. The char flows are based on ultimate char yields and elemental compositions, as devolatilization occurs in a hypothetical stand-alone process that covers both primary devolatilization and tar decomposition, either instantaneous or with finite-rate kinetics. In nonreactive carrier streams, the char simply passes through the reactor network without modification. In principle, it could provide reaction sites for gaseous species although in the current analysis, no heterogeneous chemistry on char is included. In addition to these direct flows into the CSTR, the reactor is surrounded by a pool of entrainment flows, FJ j , that release an entrainment increment, FE j , into the jth CSTR. Entrainment comprises mixing of secondary and tertiary streams, as well as all products of drying, primary devolatilization, and tar decomposition. With nonreactive gas flows, additions of secondary and tertiary streams simply dilute the coal suspension. These additions are evaluated from the entrainment laws in Eqs. (5.5a), (5.5b) for the time increment associated with the jth CSTR, summed over all entrainment sources. If this CSTR is within a network branch that represents the earliest stage of heating, such as a core region then, in principle, the entrainment pool comprises evaporating coal moisture or, perhaps, slurry moisture. However, drying is almost always finished before all other coal decomposition chemistry begins. It is easier to simply add inherent or slurry moisture to the feedstream into the first CSTR in a network, and omit the dynamics of drying from the analysis.

Fig. 5.10 Nomenclature for the species balance on a CSTR in an equivalent reactor network.

142

Process Chemistry of Coal Utilization

The most important entrainments from the coal suspension are primary volatiles and subsequent tar decomposition products. For temperatures below about 600°C, the entrainment increment into a particular CSTR is the incremental yield of primary volatiles for the time increment of the subject CSTR, which comprises CO, CO2, H2O, H2, H2S, HCN, and C1–C3 GHCs. These volatiles contain pristine primary tar which remains intact throughout its passage through the network because of the relatively cool temperatures. Consequently, the only chemical transformations occur via homogeneous reforming of noncondensables, which requires total reaction times of several seconds or longer because of the cool temperatures. At temperatures from 600°C to 800–900°C, depending on reaction times, the entrainment increment is the incremental yield of primary noncondensables plus tar decomposition products including additional noncondensables, oils, secondary tars, and PAH. As these volatiles comprise all the molecular species in primary noncondensables plus oils and PAH, the homogeneous reforming mechanism must cover oils as single-ring aromatics and, perhaps, PAH. In the analysis, PAH can be converted by simplified homogeneous reforming chemistry, or instantaneously converted into lighter hydrocarbons by a heuristic global process, or passed into the effluent without modification. For temperatures above 850°C or so, the entrainment increment is the incremental yield of primary noncondensables plus tar decomposition products including additional noncondensables, oils, secondary tars, PAH, and soot. However, at such high temperatures, oils, secondary tars, and PAH are short-lived intermediates that are ultimately incorporated into soot along with substantial amounts of GHCs. This scenario is amenable to two representations. First, tar decomposition can be regarded as instantaneous, whereby the volatiles comprise CO, CO2, H2O, H2, H2S, HCN, CH4, C2H2, and soot. These products reflect instantaneous reforming of all GHCs into CH4 and C2H2, and a global sooting process that eliminates tar-O as CO; tar-H as H2; and tar-N as HCN with simultaneous incorporation of C2H2 and one-third of tar-N into soot. Given a yield and elemental composition of primary tar, the yields of soot and secondary noncondensables can be calculated with simple algebraic relations (see Section 7.5.3 of the first volume in this series or Niksa, 2019b). As secondary pyrolysis is deemed instantaneous, each increment of primary volatiles is converted into secondary volatiles, including soot, as soon as it is released; that is, at the primary devolatilization rate. No tar decomposition kinetics are required. In the second scenario, each increment of primary volatiles begins to react as soon as it is released at rates given by the FLASHCHAIN®-based mechanism for tar decomposition (Niksa, 2017). This mechanism represents primary tar decomposition into noncondensables, oils, and PAH with subsequent nucleation of secondary tar and PAH into soot, followed by soot addition by GHCs, oils, and secondary tars. It has already been validated for temperatures through 1200°C. The finite-rate decomposition of primary tar introduces a time lag into the analysis whereby primary volatiles including tar persist for times determined by the gas temperature history and tar decomposition kinetics. Because the kinetics for primary devolatilization and tar decomposition are comparable at temperatures through 1000°C, these lags are often substantial, especially when tars decompose while the coal suspension is being rapidly heated. The secondary volatiles represented this

Simulations with detailed chemistry

143

way comprise all primary volatiles including primary tar plus all secondary volatiles including oils, PAH, secondary tar, and soot. By definition, all volatiles except soot in an entrainment increment instantaneously mix and react with the gases in a CSTR. At high temperatures, soot forms and is treated as a condensed phase species analogous to char. After the onset of soot nucleation, the analysis includes a flow of soot across the CSTR denoted by FS j and FS j + 1 . With nonreactive carrier gas, the soot is also nonreactive, unless the recombination reactions in the soot surface chemistry mechanism in Table 5.2 factor into the calculations. The conservation principle for species i across the CSTR is given by FjP+ 1 yjP,+i1  FjP yjP,i  FjE yjE, i

ð ¼ ωi Mi dV

(5.10)

where subscript i denotes a particular molecular species within any of the flows; yj K, i is its corresponding mass fraction in stream K, where K denotes either the process stream or the entrainment stream; Mi is the molecular weight of species i; ωi is the net volumetric generation rate of species i from homogeneous chemistry; and dV is a volume increment. As the composition within the jth CSTR is uniform, by definition, the integral is readily evaluated as the product of the mean homogeneous reaction rate and the reactor volume. An overall mass balance has a similar form, except that neither mass fractions nor the homogeneous reaction term appears because chemistry in the gas phase cannot change the mass of the process stream: FjP+ 1 ¼ FjP + FjE

(5.11)

Eqs. (5.10), (5.11) constitute n + 1 implicit equations for the n + 1 unknowns, FP j + 1 and yP,i j + 1 , where n is the number of species. Once specified, these unknowns determine the homogeneous reaction rate, ωi, because the homogeneous reaction mechanism specifies all production rates in terms of the species concentrations. Hence, this equation set determines the composition and mass flow rate of the process stream leaving the jth CSTR. The calculations move forward in time starting with the first CSTR. For every reactor, the inlet and entrainment flowrates and compositions are known in advance, and the outlet streams, once determined, become the inlet for the succeeding reactor. As the effluent composition from a CSTR develops instantaneously from the inlet composition and remains fixed in time, the following convergence procedure is the most expedient: (1) Specify the flowrates and compositions of all inlet streams into the jth CSTR. (2) Use the inlet compositions as an estimate for the product composition. (3) Use the estimated composition to evaluate the homogeneous chemistry in the CSTR, which automatically delivers an updated product composition. (4) Return to step 3 until the iterations converge to a stable product composition.

Usually, the gas compositions throughout a 30-CSTR reactor network are converged in under a minute on ordinary personal computers. It is expedient to add time

144

Process Chemistry of Coal Utilization

derivative terms to Eqs. (5.10), (5.11) weighted by a very small number so that robust stiff ODE solvers can be used instead of less stable solvers for systems of nonlinear algebraic equations. The following case study illustrates the impact of representing tar decomposition as instantaneous vs finite-rate under high temperature, rapid heating conditions. The subject reaction system is a single coal jet from an injector in a 550 MW T-fired utility furnace. CFD simulations based on detailed engineering specifications determined the operating conditions along the coal jet, including coal mass loading, thermal histories, and transit times. A reactor network developed from the CFD results staged the chemistry in this environment with realistic utility grinds for three diverse coal types. The chemistry simulations cover the period for primary devolatilization, tar decomposition, and secondary volatiles pyrolysis because the O2 in the actual carrier gas was omitted from this case study. The jet consists of a p. f. coal grind entrained in N2 at mass loadings approaching 0.5 kg-coal/kg-N2. Coal loadings were adjusted to maintain the same furnace firing rate for the different coal types. Once the coal suspension leaves the injector nozzle, it heats primarily by radiation from the furnace and loses heat from convection to its carrier gas. As coal jets in commercial furnaces are momentum-dominated, entrainment of surrounding air into the primary jet was negligible during this period, and there was also no impingement with other coal jets or any wall element. The mean thermal histories for primary gases and coal appear in Fig. 5.11. They were evaluated from the CFD results as mean values across the primary stream transverse to the flow direction. These histories account for heat release from volatiles combustion and char burnout by the air streams in the actual furnace but are used here with a nonreactive carrier gas simply to illustrate the secondary pyrolysis behavior under realistic furnace conditions. Both temperatures increase at two distinct heating rates. The first heating stage lasts for 130 ms, while the primary jet received energy mostly from auxiliary air streams that were preheated to 335°C. The coal suspension heated at about 2000°C/s during this stage but remained below 300°C. Consequently, this stage is regarded as nonreactive, and simply adds a 130 ms lag to the onset of chemistry in the jet. The second stage covers the remaining 210 ms, while the primary stream heats at approximately 104°C/s primarily by radiation from the macroscale coal flame. Whereas the literature contains numerous estimates for coal heating rates in utility furnaces as fast as 106°C/s, the bulk of the coal in this primary stream does not heat nearly so fast. Momentum-dominated jets with such high mass loadings cannot possibly heat as fast as individual 51 μm particles. Both the coal and carrier gas heat at similar rates although gas temperatures are 70°C cooler than coal temperatures because only the particles absorb the intense thermal radiation. The curve labeled TRAD in Fig. 5.11 is not the equivalent radiation source temperature in the CFD simulations. Instead, it is the radiant source temperature that gives the same coal particle temperature as the CFD mean values in an enthalpy balance for individual entrained particles. The particle energy balance incorporated the CFD gas temperature history, and gave the same particle temperature history as the CFD coal history for the assigned TRAD history. In other words, the particle energy balance was inverted to

Simulations with detailed chemistry

145

1800 1600

TRAD

Temperature, °C

1400

TCOAL

1200

TGAS

1000 800 600 400 200 0 0.0

51 mm 0.1

0.2

0.3

0.4

Time, s

Fig. 5.11 Mean thermal histories for (▪) the coal suspension and (dashed curve) carrier gas from a CFD simulation of the primary injection stream, compared to (solid curve) the coal thermal history for 51 μm particles from an energy balance that incorporates (dot-dashed curve) the estimated radiant source temperature and CFD carrier gas temperature. Reproduced from Niksa S. Simulating volatiles conversion in dense burning coal suspensions. Part 2. Extrapolations to commercial p. f. firing conditions. Fuel 2019c;252:832–40 with permission from Elsevier.

assign TRAD that matched the coal thermal history for the same gas thermal history. As seen in Fig. 5.11, the agreement between the coal thermal histories from CFD and the particle energy balance is nearly exact once the temperatures exceed the threshold for primary devolatilization. The subject flowfield is a premixed primary jet with no entrainment of auxiliary air, so its equivalent reactor network is simply a linear CSTR series with 12 reactors for the initial nonreactive stage and 21 reactors for the radiant heating section. A series with 33 CSTRs closely approximates plug flow. The nominal transit time from the CFD results was subdivided equally among the 33 CSTRs, which gave a nominal residence time of 10 ms in each unit. The mean thermal histories for gas and radiant source temperatures were approximated as discrete isothermal values for each CSTR in the series. Given a large number of CSTRs and the relatively short transit times, the discrete approximation to the mean CFD thermal history is virtually exact. The average gas and radiant source temperature histories were incorporated into an enthalpy balance that reproduced the thermal histories for the particles in suspension from the CFD results, as shown in Fig. 5.11. These thermal histories were then incorporated into FLASHCHAIN® to determine the complete product distributions for both primary devolatilization and instantaneous conversion of tar into soot. In a second calculation series, the FLASHCHAIN®-based tar decomposition mechanism was

146

Process Chemistry of Coal Utilization

evaluated over the thermal history of the gas to determine the incremental additions of noncondensables and soot for finite-rate secondary volatiles pyrolysis. Both calculation sweeps included tar decomposition into soot because the gas temperature exceeded the sooting threshold within 250 ms. The predicted product distributions were then rendered into discrete injections of volatiles into each CSTR in the series that matched increments in the transit times to each CSTR residence time. The first CSTR was fed with the entire char stream plus the entire primary carrier stream plus inherent coal moisture, and volatiles were injected into only those CSTRs whose temperatures were hotter than the threshold for primary devolatilization. Subbituminous, hv bituminous, and lv bituminous coals were simulated. As the bituminous coals have appreciably different calorific values from the subbituminous, the coal feed rates were adjusted to maintain the same furnace rating, whereas the primary N2 flow rate was the same for all coals. The same grind was used with each coal and had a mean size of 50.8 μm. In the simulations, the PSD was resolved into 30 discrete size increments. The predicted concentrations of the major reaction products with hv bituminous appear in Fig. 5.12 for both tar decomposition scenarios. The N-species are not shown because only HCN was present with very small amounts of NH3, and the HCN concentrations increased throughout the process. Soot levels are indicated as a mass flowrate whereby a soot yield of 30 daf wt.% corresponds to a soot flowrate of 0.7 kg/h. The differences between the two scenarios are striking at every stage of this chemistry. For instantaneous tar decomposition, sooting coincides with primary tar production, 1.50

0.8

1.50

0.8 hv bituminous

hv bituminous

Soot

1.25

CH4

1.25

0.2

C2H2 0.25 0.00 17.5 15.0

CH4

0.0 H2

0.4

0.75 0.50 C2H2

0.25

C2 H4

hv bituminous

Soot

1.00

C2H4

0.00 hv bituminous

0.2

0.0 17.5 15.0

H2

12.5

12.5 10.0

7.5

7.5

5.0 4

5.0 4

Concentration, %

10.0

3

3 2 1 0 150

200

250 Time, ms

300

H2O

H2O

CO

CO

CO2

CO2 150

200

250 Time, ms

Concentration, %

0.50

Concentration, %

0.4

0.75

Soot Flowrate, kg/s

Concentration, %

1.00

Soot Flowrate, kg/s

0.6

0.6

2 1 0

300

Fig. 5.12 Predicted products of secondary volatiles pyrolysis for (left) instantaneous and (right) finite-rate tar decomposition along a jet of hv bituminous coal.

Simulations with detailed chemistry

147

which occurs from 150 to 225 ms. As the instantaneous global soot formation process produces CO, H2, and HCN and consumes GHCs, all species concentrations evolve on a common time scale. The only exception is the moderate conversion of C2H2 into C2H4 from 200 to 250 ms, which also perturbs the CH4 concentration. For this heating rate, primary devolatilization is finished by the time the coal heats to 1000°C. Accordingly, all product concentrations achieve their ultimate values by 250 ms if tar decomposition occurs at the devolatilization rate. Volatiles reforming is negligible thereafter despite the hotter temperatures. With finite-rate tar decomposition, the volatiles concentrations develop on two distinct time scales, one for primary devolatilization and partial tar decomposition into oils and PAH and another for sooting. The first process occurs from 150 to 200 ms and determines the intermediate levels of CO, CO2, and H2O. Tar decomposition into oils and PAH releases additional GHCs which are reformed into CH4 and C2H2 at moderate temperatures in this stage. Secondary tars nucleate into soot at 250 ms, then rapidly add to soot along with GHCs, oils, and PAH thereafter. The lag between the onset of primary devolatilization and sooting is 100 ms. The concentrations of GHCs, CO2, and H2O are established by primary devolatilization, whereas those of CO, H2, and soot are substantially enhanced by sooting. Sooting is also responsible for modest reductions in the GHC levels. The effluent compositions from both tar decomposition scenarios are very similar in their concentrations of H2, oxygenated gases, GHCs, and soot. The only exception is that finite-rate tar decomposition gives five times more CH4 than instantaneous decomposition. But effluent compositions are much less important than the dynamic compositions during the initial stages of p. f. firing. The most significant difference in the dynamics is that much greater GHC concentrations form much earlier with finiterate tar decomposition, at the expense of delayed H2 production. In the presence of O2, this difference would promote a faster initial heat release and more effective NO reduction. The coal decomposition chemistry is clarified further in the gas compositions and soot flowrates for subbituminous and lv bituminous in Fig. 5.13. Both cases are for finite-rate tar decomposition. Both coals generate greater concentrations of GHCs and less soot and H2 than the hv bituminous, as expected. The subbituminous gives the greatest concentrations of oxygenated gases, by far, whereas the lv bituminous gives hardly any, consistent with its very low O-content. The onset of devolatilization is slightly advanced with the subbituminous, and retarded by 25 ms with the lv bituminous. Consequently, soot nucleation is shifted by the same extents, so the lag from devolatilization to soot nucleation is very similar with all coals. All coals also exhibit the rapid reforming of heavier GHCs into CH4 and C2H2 at moderate temperatures, as well as surges in the concentrations of CO, H2, and CH4 throughout sooting.

5.5.2 Equations for gas streams with O2 The vast majority of commercial applications of suspension firing have one or multiple gas streams that contain O2, as in furnaces and entrained flow gasifiers. These systems sustain all the chemistry for primary devolatilization and tar decomposition

148

Process Chemistry of Coal Utilization 3.0

0.8 subbituminous

2.5

3 0.4 2

Soot 0.2

1

CH4

0.6

C2H2

0.6 Concentration, %

CH4

Soot Flowrate, kg/s

4

2.0 1.5 1.0 0.2 C2H2

0.5

C2H4

C2H4 0 17.5

0.0 subbituminous

0.4

Soot

0.0

0.0 17.5

lv bituminous

15.0

15.0 12.5

12.5 Concentration, %

10.0

H2

H2

7.5

10.0 7.5

5.0 4

H2O

5.0 4

3

CO

3

Concentration, %

Concentration, %

0.8 lv bituminous

Soot Flowrate, kg/s

5

2

2 1

H2O CO

CO2

0 150

200

250 Time, ms

300

150

200

250 Time, ms

1 CO2

0

300

Fig. 5.13 Predicted products of secondary volatiles pyrolysis for finite-rate tar decomposition along jets of (left) subbituminous and (right) lv bituminous coal.

in the previous section, as well as volatiles combustion and the burnout of char and soot. The chemical reaction mechanism for homogeneous chemistry must comprise all the reactions used for secondary volatiles pyrolysis as well as many additional reactions for the oxidation of all gaseous fuel components. Moreover, this mechanism must be validated for an enormous range of SR, because turbulent mixing determines the local SR value. Even though coal suspensions are normally entrained in air, the initial SR of the carrier gas is infinite, because there are no gaseous fuel components in a primary air stream. As the suspension is heated beyond the threshold for primary devolatilization and tar decomposition, gaseous fuel components accumulate in the primary stream, as seen in Figs. 5.12 and 5.13, which diminishes local SR values when the primary stream contains O2. Once the gas reaches the ignition temperature, the gaseous fuel components are oxidized while additional gaseous fuels are added from the coal particles into the gas phase. The gaseous fuel components do not burn in isolation of char and soot (as covered in Chapter 8 of the first volume in this series). Rather, the burning rates of noncondensable fuels are strongly mediated by the competition for the available O2 among gaseous fuels, soot, and char. Consequently, volatiles burning rates in coal flames are roughly an order of magnitude slower than in gaseous fuel mixtures, even though the volatiles fuel mixtures in isolation burn as fast as synthetic GHC mixtures. Gaseous fuels are the fastest burning component in flames, but GHCs persist while substantial amounts of soot and char burn away. Burnout as a function of SR of the noncondensable fuels and soots from different coals are very similar, whereas char

Simulations with detailed chemistry

149

oxidation competes less effectively for O2 with coals of progressively higher rank. The competition from char burnout also depends on the coal PSD because burning rates accelerate for progressively smaller sizes. Elevated pressure does not directly affect the relative burnout of the different fuel components, although it usually perturbs local conditions such as gas temperatures. Contrast this explicit competition with the treatment of volatiles combustion in CFD furnace simulations. CFD concentration fields omit reaction mechanisms for volatiles conversion. Usually, no finite-rate volatiles combustion chemistry of any kind is included in CFD chemistry submodels, because this chemistry is confined to thin layers that cannot be resolved on the coarse computational grids of current furnace simulations. Instead of using even global oxidation rate expressions, volatiles combustion submodels determine the equilibrium compositions of the combustion products from the elemental compositions of volatiles with Gibbs free energy minimization and set the overall burning rates equal to turbulent mixing rates of secondary air streams into fuel-rich, primary coal jets. Such analyses do not distinguish between primary and secondary products, so they omit soot. Soot is often the slowest burning fuel component in the system and sequesters major portions of what began as GHCs, which have the fastest burning rates of all fuel components. Consequently, this omission distorts the competition for O2 among the fuel components in the analysis. The equilibrium analysis for volatiles combustion also explicitly decouples the conversion of char from gaseous fuels. This is an important omission from the mechanisms responsible for fuel-N conversion and NO production, and also for the development of nascent syngas in gasifiers. The analysis in this section rectifies these shortcomings and may, in principle, provide the clearest view of the process chemistry during the earliest stages of p. f. firing under commercial conditions. However, there is no means to stringently evaluate these findings with measurements under comparable operating conditions. Measurements to diagnose this environment are impossible now and for the foreseeable future because of the heavy particle loadings, very fast velocities, short time scales, and severe thermal histories. Whereas the combustible material in char is mostly carbon, the char in p. f. suspensions contains nonnegligible amounts of hydrogen, very significant amounts of nitrogen, and, perhaps, trace amounts of sulfur and oxygen. The conversion of char-H into H2O and char-S into SO2 are unambiguous, and one can safely assume that char-O is released as O2 because the amount being converted is inconsequential. Conversely, char-N conversion is so complex that it is best to adopt the simplest modeling strategy. Char-N is converted at the rate of char burnout (from CBK/E), and a portion is released as NO while the remainder contributes to N2. The NO fraction, FNO, is fixed throughout all stages of burnout. This leaves the primary product of carbon oxidation to consider in more detail. When char particles in the p. f. size grade burn, the primary product of the carbon oxidation is CO. The CO is then subsequently oxidized to CO2 in the gas phase as it is transported into the free stream. It is not clear how far from the char surface the CO burns, except that this distance shrinks with larger particles. On this basis, both CO and CO2 are designated as carbon oxidation products from the char surface.

150

Process Chemistry of Coal Utilization

The fraction of carbon converted into CO, FCO, is a function of temperature that is evaluated for each size increment in the char PSD, because different sizes have different temperatures. Based on these assignments of the char oxidation products, the moles of O2 required per mole of combustibles in char is defined as υC,O2 ,O2 ¼ xC


 FCO xH xO xN + ð1  FCO Þ +  + FNO + xS 2 4 2 2

(5.12)

where the xi are mole fractions of the five elements in char combustibles. The individual terms in this equation are the molar stoichiometries for all char oxidation products except N2, whose coefficient is (1  FNO)xN/2. The coefficient for O2 is negative, whereas all others are positive. The analogous stoichiometry for soot burnout, νS, O2, O2 has FCO set to unity and omits the terms for O and S. The species conservation laws for burning coal suspensions are distinctive for the molecular species affected by the burnout of char and soot. For the unaffected species, the species balances are the same as Eq. (5.10). For O2, CO, CO2, H2O, NO, N2, and SO2, the balances are +1 FjP+ 1 yjP,i  FjP yjP,i  FjE yjE,i

  j υC,i,O2 ΔXC,O F0 1  y0A Mi 2 C ¼ ωi Mi dV + MC’ j ∞ υS,i,O2 ΔXS,O2 FS Mi + MS’ ð

(5.13)

where the term for char oxidation contains the incremental burnout predicted by CBK/E for the residence time increment of the jth CSTR, ΔXjC, O2 and the flowrate of ash-free combustibles into the first CSTR in the series, FC 0 ð1  yA 0 Þ. The stoichiometric coefficient νC, i, O2, and the mean char molecular weight, MC0 , account for the presence of heteroatoms in the char combustibles. The term for soot oxidation contains the burnout predicted by NSC kinetics and the flowrate based on the ultimate yield of soot for the most severe conditions in the near-injector region. This basis ensures that the extent of soot burnout varies from zero to unity regardless of the fractional soot flowrate into any particular CSTR. The stoichiometric coefficients and the mean soot molecular weight, MS0 , account for the presence of H and N in coalderived soots. An overall mass balance has a similar form, except that the burnout terms are weighted by the net gain of mass into the gas phase, as in FjP+ 1

  8 j 0 0 X ΔXC, O2 FC 1  yA + υC,i, O2 Mi MC’ i¼1 j 5 ΔXS,O F∞ X 2 S + υS,i,O2 Mi MS’ i¼1

¼ FjP

+ FjE

(5.14)

Simulations with detailed chemistry

151

As the composition within a CSTR is uniform and equal to the outlet composition, the only unknowns in the species balances in Eq. (5.13) are the outlet mass fractions of species i. In other words, given a value for yP,i j + 1 , CBK/E and the NSC burning rate expression can be used to evaluate the incremental extents of burnout for char and soot, and the homogeneous reaction mechanism can be used to evaluate the species generation rate from the combustion of gaseous fuels. This analysis does not determine the apportioning of O2 among the various fuels in this reaction system in advance or through any imposed constraints. Finite-rate kinetics for oxidation of soot, char, and gaseous fuels govern O2 apportioning, as in actual pc flames. This feature is probably the strongest advantage of an equivalent network, because comprehensive reaction mechanisms under commercial operating conditions are currently only feasible within this framework. Extents of burnout for char are determined with CBK/E simulations from the ignition time up to the transit time for a current iteration of Eqs. (5.13), (5.14). Each simulation progresses through the histories of O2 concentration and ambient temperatures throughout the char burnout history. Char temperatures are not assigned as the gas temperatures in each CSTR. Instead, the particle thermal history is based on an enthalpy balance that accounts for the thermal capacitance, convection to the gas stream, radiant transfer from the surroundings, and the heat release due to char burnout (Niksa et al., 2003). CBK/E evaluates annealing rates that diminish char burning rates for progressively hotter char temperatures, which often causes extinction. The initial char reactivity is responsible for ignition and the onset of the competition for O2 with the other fuel components. But immediately after ignition, annealing comes into play while the char temperature surges and the burning rate becomes limited by film diffusion. Annealing is especially important at the hottest char temperatures and often causes extinction from diffusion-limited combustion to a kinetically limited, slowburn state during the latest stages of burnout. This impetus for extinction may be compounded by ash encapsulation. Char PSDs are represented with Rosin-Rammler PSDs and subdivided into at least 30 size increments. Each size is simulated separately so that the total burnout increment for a particular CSTR is a summation of the increments for each size weighted by the char PSD. The basis for this approach is that char moves through a CSTR in plug flow, rather than with the exit age distribution of the gas flows. This assumption avoids the highly unwieldy situation where char of the same size has widely variable extents of burnout due to the age distributions of CSTRs in the network. Instead, only a single extent of burnout pertains to each size increment. However, these extents of burnout are highly variable across the char PSD due to shifts in the rate-limiting mechanisms in CBK/E when different sizes are exposed to the same ambient conditions. Extents of burnout for soot are determined in the same way, except that soot has only one initial size of 0.8 μm, and the burning rate is based on NSC kinetics. Stable simulations of Eqs. (5.13), (5.14) have been obtained with the following procedure: First, the feasible range of ΔXC, O2j is evaluated from zero to the burnout associated with the inlet value of yj P,i , omitting all conversion of gaseous fuels and soot. This range is input into a root finder based on Brent’s algorithm. At each step in this iteration, the inlet gas composition is first adjusted for the consumption of O2 and

152

Process Chemistry of Coal Utilization

production of CO, CO2, H2O, SO2, NO, and N2 that are associated with the trial burnout level. This fictitious gas composition is then entered into simulations for the simultaneous oxidation of gaseous fuels and soot. Trial values are converged in the second application of Brent’s algorithm. The outlet O2 level from the chemistry simulation is used in another char oxidation run with CBK/E. When the updated value for the incremental burnout level matches the trial value, the CSTR simulation satisfies the O2 balance in Eq. (5.13), and the calculation sequence moves forward to the next CSTR in the series. Simulations with 40 CSTRs generally converge for the relevant range of SR values in actual pc flames. Each simulation with a series of 40 CSTRs takes a few minutes on an ordinary microprocessor. The archetypal case study for coal suspensions that contain O2 is the same reaction system covered for nonreacting gas streams, except that the N2 carrier gas is replaced by air. The subject reaction system is a single primary coal jet from an injector in a 550 MW T-fired utility furnace. The CFD simulations that specified coal mass loading, thermal histories, and transit times for the nonreacting case were actually run with air streams. So the previous reactor network and the thermal histories in Fig. 5.11 are even more suitable for the coal jet in a primary air stream. The chemistry simulations cover the period for primary devolatilization, tar decomposition, and secondary volatiles pyrolysis, volatiles combustion, char particle ignition, and the partial burnout of char and soot. This case was already covered in detail in Section 8.6.2 of the first volume in this series and by Niksa (2019c). The main features of the flame structure through 340 ms are considered here, along with the impacts of loading variations, the p. f. grind, and coal quality on NO production and char burnout. The predicted flame structure in Fig. 5.14 is for a hv bituminous with 0.35 kg-coal/kg-air. This loading is slightly lower than the commercial loading of 0.42 kg-coal/kg-air but both flame structures are very similar. The flame structure conveys three distinct conversion time scales. The fastest burning fuels—C2H2, CH4, and H2—appear to be converted from 140 to 200 ms. However, only about three-fourths of these fuels were released from the coal during this period, as volatiles continued to be released through 270 ms. GHCs continue to burn away or are reformed into CO and H2 through most of the total transit time. The second major stage from 180 to 210 ms oxidizes away CO; converts HCN into NO via an N2O intermediate; and burns out some char, while almost all O2 is consumed. The bulk of the HCN forms N2 rather than NO although this case is so reducing that onethird of the maximum HCN concentration remains as HCN in the effluent. The third and slowest conversion period extends from 210 to 325 ms, when soot is converted under very low O2 concentrations and the last portion of volatiles is added to the entrainment stream to sustain further NO reduction. The CO product released from volatiles reforming and soot combustion promotes water gas shifting, which elevates the H2 concentration. In turn, H2 and CO convert most of the NO back into HCN along the postflame region. With this particular coal and coal loading, char oxidation coincides with the oxidation of all gaseous fuels and N-species conversion, whereas soot oxidation in conjunction with volatiles reforming promotes water gas shifting chemistry, CO accumulation, and major extents of NO reduction along the postflame region.

Simulations with detailed chemistry

hv bituminous 0.35 kg-coal/kg-air

hv bituminous 0.35 kg-coal/kg-air 0.75

5 Char

4

0.50

3 2

Concentration, %

0 22.5 20.0 17.5 15.0 12.5 10.0 7.5 5.0 2.5 2.0 1.5 1.0 0.5

0.25

C2H2

Soot

1

Concentration, %

Burnout, %

6

CH4 0.00 O2

CO2

hv bituminous 0.35 kg-coal/kg-air

1750 1500

H2O

1250 1000 HCN

750 500

CO hv bituminous 0.35 kg-coal/kg-air

H2

0.0 150

200

250 Time, ms

300

350

NO

N2O 150

200

250 Time, ms

300

Concentration, ppm

7

153

250 0 350

Fig. 5.14 Predicted flame structure with hv bituminous at 0.35 kg-coal/kg-air as, in clockwise order from the upper left panel, extents of burnout of char and soot; the concentrations of CH4 and C2H2; HCN, N2O, and NO; and O2 and major gas intermediates (H2, CO) and products (CO2, H2O). Reproduced from Niksa S. Simulating volatiles conversion in dense burning coal suspensions. Part 2. Extrapolations to commercial p. f. firing conditions. Fuel 2019c;252:832–40 with permission from Elsevier.

The postflame is sufficiently reducing to regenerate CH4 and C2H2 in tandem with NO reduction. Consequently, volatiles combustion does not completely consume even the most reactive gaseous fuels. However, the simulations did not account for the incorporation of C2H2 into soot, which should occur throughout the postflame, so the C2H2 concentration in Fig. 5.14 is probably too high. The impact of loading variations on the flame structure is apparent in Fig. 5.15 for subbituminous coal. With lighter loadings, the suspension releases less gaseous fuels, but the dynamics of volatiles combustion are otherwise the same with all loadings. Char ignition is also the same and, with this coal, char oxidation always precedes soot oxidation by 50 ms. With subbituminous, extents of char burnout are always substantial, increasing from 17% with the heaviest loading to 82% with the lightest loading. However, soot burnout is only appreciable with the lightest loading, for which it is only one-third of char burnout. Extensive char burnout coincides with volatiles combustion, which consumes O2 even before all the gaseous fuels have burned away. The key shift in the flame structure is that O2 is eliminated faster with progressively heavier loadings. With 0.15 kg-coal/kg-air, over 80% of char and a quarter of soot burn out yet O2 persists to 300 ms in the gas stream. Oxygen is eliminated at shorter transit times with progressively heavier loadings.

154

Process Chemistry of Coal Utilization 2500 HCN

0.15

80

Burnout, %

60

40

Char

0.35

2000

0.35

1500

1000

0.15

0.15

0.46 Soot

0

0.46

0.25

0.25

20

subbituminous

150

200

250 Time, ms

500

NO

0.15

Concentration, ppm

subbituminous

0.25 300

350

150

200

250

300

0 350

Time, ms

Fig. 5.15 Predicted combustion histories with subbituminous at four loadings showing (left) extents of burnout of (solid curves) char and (dashed curves) soot; and (right) molar concentrations of (dashed curves) HCN and (solid curves) NO. The two lowest NO curves correspond to 0.35 and 0.46 kg-coal/kg-air, respectively. Reproduced from Niksa S. Simulating volatiles conversion in dense burning coal suspensions. Part 2. Extrapolations to commercial p. f. firing conditions. Fuel 2019c;252:832–40 with permission from Elsevier.

Consequently, as seen in Fig. 5.15, NO reduction remains strong throughout the postflame with the three heavier loadings, and effluent NO levels become much lower than the maximum levels at the end of volatiles combustion. The subbituminous generates much lower near-burner NO than the hv bituminous for two reasons: First, the maximum NO levels at the end of volatiles combustion are lower because the coal-N is lower and also because HCN destruction skews toward N2 with progressively lower O2 concentrations; and, second, NO reduction along the postflame is stronger due to the greater H2 and CO concentrations associated with greater extents of char burnout. With this coal, only the lightest loading gives an effluent NO level comparable to the maximum at the end of volatiles combustion because persistent O2 suppresses NO reduction. Clearly, the persistence of O2 with both lighter loadings suppresses NO reduction along the postflame region, because H2 and CO reductants cannot coexist with O2 at these temperatures even though both char and soot burn out along the entire postflame with lighter loadings. Conversely, the elimination of O2 allows the gaseous reductants to accumulate along the postflame and drive the effluent NO concentration well below its maximum, even when extents of char and soot burnout are negligible. The impact of char oxidation is strongly affected by the coal PSD, as evident in Fig. 5.16, which shows burnout as a function of initial coal size across the grind for subbituminous at four coal loadings. With all loadings, the general tendency is for diminishing extents of burnout for progressively larger sizes, although with the three heavier loadings, burnout passes through a maximum at roughly 60 μm. Burning rates for the smallest sizes are limited by their rapid heat loss rates (which are inversely proportional to size), which keep their temperatures close to the gas temperature. Intermediate sizes provide the best balance between the rates of heat losses and

Simulations with detailed chemistry

155

100 Cum. PSD

Char BO & Cum. PSD, %

80 0.15 60 subbituminous 40

0.25 0.35

20

0

0.46

0

50

100

150

200

250

Particle Size, mm

Fig. 5.16 (Dashed curve) Cumulative Rosin-Rammler PSD for the coal grind; and (solid curves) predicted char burnout vs initial coal size for four loadings of subbituminous coal. Reproduced from Niksa S. Simulating volatiles conversion in dense burning coal suspensions. Part 2. Extrapolations to commercial p. f. firing conditions. Fuel 2019c;252:832–40 with permission from Elsevier.

O2 transport and give the greatest excursions above the local gas temperature. The burning rates of the largest sizes are slower because O2 transport rates diminish for progressively larger sizes. With the lightest loading, burnout varies from 20% to almost 95% across the PSD. Whereas this size dependence is clearly an essential aspect of the competition for O2 among the fuel components, grind specifications for pc firing do not vary very much with different coals. With lv bituminous (Niksa, 2019c), char and soot ignite at comparable transit times and soot burns much faster than char. More than half the soot burns out with the lighter loadings, but char burnout never exceeds 20%. Most important, the slow heterogeneous burning rates are responsible for sequential volatiles combustion and soot burnout, so that NO production is unaffected by heterogeneous combustion with both intermediate loadings. With the heaviest loading, NO diminishes from its maximum at the end of volatiles combustion, albeit by only one-third. With the lightest loading, the maximum NO concentration is the lowest of all because the accumulation of HCN is at a minimum. Based on these connections among the sequence of conversion of the different fuel components and NO production, the concentration histories for H2, CO, and O2 with the subbituminous and lv bituminous coals in Fig. 5.17 illustrate how coal quality affects near-burner NO production. In the most favorable situation, as with the subbituminous, char burns fast enough to accelerate O2 consumption during volatiles combustion. Rapid O2 consumption biases the partitioning of HCN toward N2, away from NO toward the later stages of volatiles combustion, which diminishes the

Process Chemistry of Coal Utilization

Concentration, %

20

O2

Iv bituminous

O2

subbituminous

15

0.15

10

0.25 CO

0.46 0.35

5

0

H2

CO

150

200

250 Time, ms

300

350

150

200

250 Time, ms

5 4 3 0.40 2 0.35 1

0.35 0.40

H2

22.5 20.0 17.5 15.0 12.5 10.0 7.5

300

Concentration, %

156

0 350

Fig. 5.17 Predicted molar concentration histories of (solid curves) CO, (dashed curves) H2, and (dotted curves) O2 with four loadings of (left) subbituminous and (right) lv bituminous. Reproduced from Niksa S. Simulating volatiles conversion in dense burning coal suspensions. Part 2. Extrapolations to commercial p. f. firing conditions. Fuel 2019c;252:832–40 with permission from Elsevier.

maximum NO concentration. It also allows H2 and CO reductants to accumulate along the postflame region. In turn, these species reduce NO concentrations to well below their maximum levels at the end of volatiles combustion. In the least favorable situation, as with the lv bituminous, O2 concentrations remain elevated throughout volatiles combustion, which skews HCN decomposition toward NO production and thereby raises maximum NO concentrations. NO concentrations will remain elevated unless the postflame can sustain H2 and CO reductants. With the two lightest loadings of the lv bituminous in Fig. 5.17, O2 persists in the effluent so that reductants are present only with the heaviest loadings. But even with the heaviest loading, no reductants emerge in the postflame region until 250 ms, so the extent of NO reduction along the postflame is relatively small. In summary, volatiles combustion under commercial loadings in an isolated primary coal suspension alters the species concentrations but does not necessarily add or eliminate any fuel components or N-species. With all but low volatility coals, the effluents contain CH4, C2H2, H2, CO, HCN, and H2S, char, and soot, although the predicted C2H2 levels are uncertain due to omission of addition to soot along the postflame region. With low volatility coals, effluents have minimal or no GHCs and appreciable amounts of NO. Soot burnout is relatively low with all coals, and char burnout is appreciable with only low-rank coals. Variations in loading do not change the dynamics of volatiles combustion or char ignition but progressively heavier loadings accelerate O2 elimination. The elimination of O2 allows the gaseous reductants to accumulate along the postflame and drive the effluent NO concentration well below its maximum, even when extents of char and soot burnout are negligible. Coal quality primarily affects near-burner NO production through variations in char oxidation reactivities. Rapid char burning rates accelerate O2 consumption during volatiles combustion, which biases the partitioning of HCN toward N2, away from NO. It also allows H2 and CO reductants to accumulate and reduce NO concentrations along the postflame region.

Simulations with detailed chemistry

157

5.5.3 Equations for gas streams with steam, CO2, H2, and CO The streams injected into entrained flow gasifiers contain O2 as well as steam and, perhaps, recycled syngas with CO2, H2, and CO. Obviously, the reaction system is more complex than coal jets entrained in air, but not because the competition for O2 by the various fuel components is expanded into a competition for O2 and the gasification agents, CO2, and steam. Oxidation is so much faster than volatiles reforming and char gasification that as long as O2 is present at levels down to 300–500 ppm, reforming chemistry and char gasification can be omitted from the analysis. In other words, the initial stages of entrained flow gasification should be analyzed as a pure combustion system until O2 levels diminish to the threshold for competitive volatiles reforming and the gasification of char and soot. The combustion behavior is attenuated by the greater suspension loadings fed into gasifiers, different grind size distributions, higher pressures, and hotter process temperatures. But the sequence of coal conversion processes during the initial stages is the same: primary devolatilization and tar decomposition, as well as volatiles combustion and the partial burnout of char and soot. Indeed, O2 consumption rates are again determined by the kinetic competition among all the various fuel components for the available O2. The additional complexity comes into play once the O2 has been nearly eliminated. The main complication in gasification systems is that levels of neither the gasification agents—steam, CO2, and H2—nor the H2 and CO inhibitors are determined by injection conditions. Rather, all these concentrations are determined by the initial oxidation chemistry with attenuation by continuous reforming of the gas phase species, and continuous production via gasification of char and soot. Homogeneous reforming chemistry moves the levels of all four gas species toward equilibration of the water gas shift reaction (CO + H2O $ CO2 + H2) until the syngas temperature is quenched. It also produces GHCs, especially CH4, at moderate gasification temperatures. Steam gasification of char and soot produces CO and H2 which inhibit the overall gasification rate, and CO2 gasification produces even more CO. Hydrogasification consumes the H2 inhibitor for steam gasification. All this chemistry occurs at pressures as high as 10 MPa and temperatures as hot as 2500°C. Only the total reaction times of 2–5 s in entrained flow gasifiers are comparable to those in pc furnaces. Another complication pertains to the coal quality impacts. The primary performance specification for any gasifier is to generate syngas compositions within tight specifications from whatever coal is being processed. This is met by adjusting O2 and steam injection rates to impose uniform overall H/C and O/C ratios with every coal sample. Consequently, suspension loadings and the gas stream injection rates are adjusted for each coal, which broadens the operating domain of the gasifier. The chemical reaction mechanism for homogeneous chemistry must comprise all the reactions for secondary volatiles pyrolysis and reforming as well as the oxidation of all gaseous fuel components. This mechanism must be validated for the elevated operating pressures and an enormous range of SR. Local SR values traverse the stoichiometric condition as the oxidizing environment in the injected coal jet gives way to deeply reducing conditions further downstream, and turbulent mixing determines the local SR value. Moreover, coal suspensions may be entrained in air, O2, or inert carrier

158

Process Chemistry of Coal Utilization

gas. Regardless, the initial SR of the carrier gas is infinite, because there are no gaseous fuel components in a primary coal stream. Further downstream, local SR levels develop in ways determined by the primary gas composition, injector configurations, and turbulent mixing rates. As the suspension is heated beyond the threshold for primary devolatilization and tar decomposition, gaseous fuel components accumulate in the primary stream, and local SR values diminish. Once the gas reaches the ignition temperature, the gaseous fuel components are oxidized, while additional gaseous fuels are added from char and soot into the gas phase by both oxidation and gasification chemistry. Ultimately, the reaction system relaxes to continuous gasification of char and soot by steam and CO2 with inhibition by H2 and CO, and equilibration of water gas shifting along the exit gas temperature profile. As in CFD furnace simulations, gasifier CFD simulations omit reaction mechanisms for volatiles conversion, with the same adverse consequences. Instead of using even global oxidation rate expressions, gasifier combustion submodels determine only the equilibrium combustion products and equate the overall burning rates to turbulent mixing rates of secondary air streams into primary coal jets. Such analyses do not distinguish primary and secondary products, so they omit soot. Soot is even more likely to appear in syngas than in flue gas. So this omission distorts the competition for O2 among the fuel components in the analysis, and may also distort predicted carbon conversions. The equilibrium analysis for volatiles combustion also explicitly decouples the conversion of char from gaseous fuels. This is an important omission from the mechanisms responsible for the development of nascent syngas in gasifiers. The analysis in this section rectifies these shortcomings, but there are no means to stringently evaluate these findings with measurements under comparable operating conditions. Measurements to diagnose this environment are impossible now and for the foreseeable future because of the heavy particle loadings, very fast velocities, short time scales, severe thermal histories, and elevated pressures. Coal grinds used in entrained flow gasifiers may be both finer and coarser than typical p. f. grinds. In O2-fired entrained flow gasifiers, grinds are finer with mean sizes under 45 μm. But when coal-water slurries are injected, grinds are much coarser and broader than standard utility grind with a mean size of at least 200 μm, as needed to stabilize the slurry viscosity. Even though the combustible material in char is mostly carbon, the char in p. f. suspensions contains nonnegligible amounts of hydrogen, significant amounts of nitrogen, and, perhaps, trace amounts of sulfur and oxygen. Whereas the major products of char gasification and hydrogasification are known, the conversion of char heteroatoms is ambiguous and rarely discussed in the literature. The fate of heteroatoms is not that important for entrained flow gasification, because the very hot temperatures eliminate most of the O and S before char conversion. But they become more important for progressively cooler temperatures. One saving grace is that homogeneous chemistry will shift any of the products of char gasification into the ultimate syngas species, provided the gasifier has sufficient transit time. The gasification stoichiometry is resolved for each gasification reaction, as follows. FLASHCHAIN® specifies the mass percentages of C, H, O, N, and S in char, which determine the molar ratios xI, where I ¼ H, O, N, S, and the denominator is

Simulations with detailed chemistry

159

always the moles of C. In terms of these molar ratios, the stoichiometry for the steam gasification chemistry (per mole of C) is rendered from C + H2 O ! CO + H2

(5.15a)

xH H !

xH H2 2

(5.15b)

xO O !

xO O2 2

(5.15c)


 1 3xN O2 xN CO + N + H2 O ! xN HCN + 2 4 xS ðS + H2 OÞ ! xS H2 S +

(5.15d)

xS O2 2

(5.15e)

In these expressions, H, O, N, and S denote moles of these elements in char. Char-H and -O form their bimolecular gases; char-N forms HCN, and char-S forms H2S. Hence, the overall steam gasification process is: h i h xN xH i Char + 1 + + xS H2 O ! ½1  xN CO + 1 + H2 2  2 xO 3xN xS + + + O2 + xN HCN + xS H2 S 2 4 2

(5.16a)

which gives the final stoichiometry once O2 is eliminated via the production of water: Char + ½1  xN  xO H2 O ! ½1 xN CO xH 3xN  xS H2 + xN HCN + 1 +  xO  2 2 + xS H 2 S

(5.16b)

Eq. (5.16b) specifies the overall stoichiometry, which also relates the incremental mass production of CO and H2 and consumption of H2O to the extent of char conversion, ΔXjC, H2O, as follows: "

# j %CC YC0 ΔXC, H2 O ΔmCO,H2 O ¼ 28 ð1  xN Þ 12

(5.17a)

"


# j %CC YC0 ΔXC,H x 3x H N O 2 1 +  xO   xS ΔmH2 ,H2 O ¼ 2 12 2 2 "

j %CC YC0 ΔXC, H2 O ð1  xN  xO Þ ΔmH2 O,H2 O ¼ 18 12

(5.17b)

# (5.17c)

160

Process Chemistry of Coal Utilization

"

j %CC YC0 ΔXC,H 2O xN ΔmHCN,H2 O ¼ 27 12

#

"

j %CC YC0 ΔXC,H 2O xS ΔmH2 S,H2 O ¼ 34:04 12

(5.17d) # (5.17e)

where %CC and Y 0 C are the carbon content and yield (in daf wt.%) of the initial char, respectively. Eqs. (5.17a)–(5.17e) determine the incremental production and consumption of gases associated with steam gasification. The amounts of CO, H2, HCN, and H2S associated with the incremental char conversions along the reactor are added to the syngas in each CSTR, while the appropriate amount of steam is eliminated. The stoichiometry for CO2 gasification chemistry (per mole of C) is rendered from C + CO2 ! 2CO

(5.18a)

xH H !

xH H2 2

(5.18b)

xO O !

xO O2 2

(5.18c)


 1 xN CO2 + N + H2 ! xN HCN + O2 2

(5.18d)

xS ðS + H2 Þ ! xS H2 S

(5.18e)

Char-H and -O form their bimolecular gases; char-N forms HCN, and char-S forms H2S. Hence, the overall CO2 gasification process is: 

5xN xH + xS  Char + ½1 + xN CO2 + H2 2 2 ! 2CO + ½xO + 2xN H2 O + xN HCN + xS H2 S

(5.19)

Eq. (5.19) specifies the overall stoichiometry, which also relates the incremental mass production of CO and H2 and consumption of CO2 to the extent of char conversion, ΔXC j , as follows: " # j %CC YC0 ΔXC,CO 2 ΔmCO2 ,CO2 ¼ 44 ð1 + x N Þ 12 "

j %CC YC0 ΔXC,CO 2 ΔmCO,CO2 ¼ 28 2 12

(5.20a)

# (5.20b)

Simulations with detailed chemistry

161

"


# j %CC YC0 ΔXC,CO 5x x N H 2 + xS  ΔmH2 ,CO2 ¼ 2 12 2 2 "

j %CC YC0 ΔXC,CO 2 ð2xN + xO Þ ΔmH2 O,CO2 ¼ 18 12

"

j %CC YC0 ΔXC,CO 2 xN ΔmHCN,CO2 ¼ 27 12

"

(5.20c)

# (5.20d)

#

j %CC YC0 ΔXC, CO2 xS ΔmH2 S,CO2 ¼ 34:04 12

(5.20e) # (5.20f)

Eqs. (5.20a)–(5.20f) determine the incremental production and consumption of gases associated with CO2 gasification. The amounts of CO, H2O, HCN, and H2S associated with the incremental char conversions along the reactor are added to the syngas in each CSTR, while the appropriate amounts of CO2 and H2 are eliminated. The stoichiometry for the hydrogasification chemistry (per mole of C) is rendered from C + 2H2 ! CH4 xH H !

xH H2 2

xO O + xO H2 ! xO H2 O


(5.21a) (5.21b) (5.21c)



3 xN N + H2 ! xN NH3 2

(5.21d)

xS ðS + H2 Þ ! xS H2 S

(5.21e)

The overall hydrogasification process is: 

3xN xH Char + 2 + + xS + xO  H2 ! CH4 + xO H2 O + xN NH3 + xS H2 S 2 2

(5.22)

Eq. (5.22) specifies the overall stoichiometry, which also relates the incremental mass production of CH4, NH3, and H2S and consumption of H2 to the extent of char conversion, ΔXjC, H2 as follows: " # j %CC YC0 ΔXC,H 2 ΔmCH4 ,H2 ¼ 16 12

(5.23a)

162

Process Chemistry of Coal Utilization

"

j %CC YC0 ΔXC,H 2 xN ΔmNH3 , H2 ¼ 17 12

# (5.23b)

"

j %CC YC0 ΔXC, H2 xS ΔmH2 S,H2 ¼ 34:04 12

#


# j %CC YC0 ΔXC,H 3xN xH 2 2+ + xS + xO  ΔmH2 ,H2 ¼ 2 12 2 2

(5.23c)

"

"

j %CC YC0 ΔXC,H 2 xO ΔmH2 O, H2 ¼ 18 12

(5.23d)

# (5.23e)

Eqs. (5.23a)–(5.23e) determine the incremental production and consumption of gases associated with hydrogasification. The amounts of CH4, H2O, NH3, and H2S associated with the incremental char conversions along the reactor are added to the syngas, while the appropriate amount of H2 is eliminated. The species conservation laws for coal suspensions in a gasifier are distinctive for the molecular species affected by the burnout of char and soot. For the unaffected species, the species balances are the same as Eq. (5.10). For O2, CO, CO2, H2O, H2, CH4, NO, N2, NH3, HCN, H2S, and SO2, the balances are ð +1 FjP+ 1 yjP,i  FjP yjP,i  FjE yjE,i ¼ ωi Mi dV   nRXN F0C 1  y0A X j υC,i,k ΔXC,k Mi + MC0 k¼1 nRXN F∞ X j + S0 υS,i,k ΔXS,k Mi MS k¼1

(5.24)

where the term for char conversion contains the burnout predicted by CBK/E for O2 levels greater than 300–500 ppm, ΔXjC,O2, and additional conversions from CBK/G for gasification by steam, CO2, and H2, ΔXjC,k, for the residence time increment of the jth CSTR. The flowrate of ash-free combustibles in char into the first CSTR in the series, FC 0 ð1  yA 0 Þ, also appears. The stoichiometric coefficient, νC,i,k, for species i in reaction k and the mean char molecular weight, MC0 , account for the presence of heteroatoms in the char combustibles. The term for soot conversion contains the burnout predicted by NSC kinetics and conversions from CBK/G for the gasification reactions; and the flow rate based on the ultimate yield of soot for the most severe conditions in the near-injector region. The stoichiometric coefficients and the mean soot molecular weight, MS0 , account for the presence of H and N in coalderived soots.

Simulations with detailed chemistry

163

An overall mass balance has a similar form, except that the heterogeneous conversion terms are weighted by the net gain of mass into the gas phase, as in FjP+ 1

  nRXN 8 X F0C 1  y0A X j + ΔX υC,i,k Mi C, k MC’ i¼1 k¼1 nRXN 5 X F∞ X j + S’ ΔXS,k υS,i, k Mi MS k¼1 i¼1

¼ FjP

+ FjE

(5.25)

As the composition within a CSTR is uniform and equal to the outlet composition, the only unknowns in the species balances in Eq. (5.24) are the outlet mass fractions of species i. In other words, given a value for yP,i j + 1 , CBK/E, CBK/G, and the NSC conversion rate expression can be used to evaluate the incremental extents of conversion for char and soot, and the homogeneous reaction mechanism can be used to evaluate the species generation rate by combustion of gaseous fuels. This analysis does not determine the apportioning of O2 and gasification agents among the various fuels in this reaction system in advance or through any imposed constraints. The kinetics for oxidation, reforming, and gasification of soot, char, and gaseous fuels govern their apportioning, as in actual gasifier streams. This feature is probably the strongest advantage of an equivalent network, because comprehensive reaction mechanisms under commercial operating conditions are currently only feasible within this framework. Extents of conversion for char are determined with CBK/E and CBK/G simulations from the ignition time up to the transit time for a current iteration of Eqs. (5.24), (5.25). Due to the strong impact of thermal annealing, every char gasification simulation must pass through the oxidation portion of a conversion history, because these portions usually have the hottest char temperatures. Each simulation progresses through the histories of O2, steam, CO2, H2, and CO concentrations and ambient temperatures throughout the char conversion history. Char temperatures are assigned from enthalpy balances that account for the heat release due to char burnout and heat absorption, depending on the stage of char conversion (Niksa et al., 2003; Liu and Niksa, 2004). CBK/E and CBK/G evaluate annealing rates that mediate char conversion rates for progressively hotter char temperatures. Annealing is especially important at the hottest char temperatures and often reduces gasification reactivities by enough to be the main cause of unconverted char in syngas. As for char burnout, char PSDs are represented with Rosin-Rammler PSDs, and subdivided into at least 30 size increments, although size variations are much less important during gasification due to the much slower reactivities. Extents of conversion for soot are determined in the same way, except that soot initially has only one size of 0.8 μm. Stable simulations of Eqs. (5.24), (5.25) have been obtained with the following procedure: For the initial oxidation stage, the protocol in the previous section is used without modification. For the gasification stage, a feasible range of ΔXjC, k is evaluated from zero to the conversion associated with the inlet values of the gasification agents, yj P,i , omitting all conversion of gaseous fuels and soot. This range is input into a root

164

Process Chemistry of Coal Utilization

finder based on Brent’s algorithm. At each step in the iteration, the inlet gas composition is first adjusted for the net consumption of the gasification agents and inhibitors that are associated with the trial burnout level. This fictitious gas composition is then entered into simulations for the simultaneous reforming of gaseous fuels and gasification of soot. Trial values are converged in the second application of Brent’s algorithm. The outlet levels of the gasification species from the chemistry simulation are used in another char conversion run with CBK/G. When the updated value for the incremental conversion level matches the trial value, the CSTR simulation satisfies the species conservation balances in Eq. (5.24), and the calculation sequence moves forward to the next CSTR in the series. Simulations with 40 CSTRs generally converge for the relevant operating domain of entrained flow gasifiers. Each simulation with a network of 40 CSTRs takes several minutes on an ordinary microprocessor. The archetypal case study for coal suspensions injected into entrained flow gasifiers is, again, one of the primary coal jets into a gasifier throughout the period for primary devolatilization, tar decomposition, volatiles combustion, and partial oxidation of char and soot. Two sets of operating conditions are considered, one for Shell gasifier conditions and another for General Electric Power Systems (GEPS) operating conditions. The jet consists of pulverized coal entrained in gas or water at mass loadings approaching 1.0 kg-coal/kg-gas and 2.0 kg-coal/kg-water. In Shell gasifiers, coal in very fine grinds is entrained in N2 or CO2 and co-fed with O2 through four nozzles in an injection plane near the base of the reactor. The nozzles are offset from the reactor radius, so that the horizontal coal jets impinge into a rectangle that rotates the flow upward in a helical pattern (Cao et al., 2018), like the flowfield in T-fired utility furnaces. In GEPS gasifiers, a coal-water slurry is fed downward through a cylindrical annulus into the reactor vessel. The coal jet is co-fed with a swirled O2 stream that mixes into the coal jet after the water evaporates. In both designs, once the coal suspension leaves the injector nozzle, it heats primarily by radiation from the coal flame, and from the heat released from fuels expelled from the coal. As coal jets in commercial gasifiers are momentumdominated, entrainment of surrounding syngas into the primary jets is deemed to be minimal during the period of primary coal decomposition. This case was already covered in detail in Section 8.6.3 of the first volume in this series and by Niksa (2019d). The main features of the flame structure through 340 ms are considered here, along with the impacts of loading variations, the p. f. grind, and coal quality on carbon conversion and syngas composition. Estimated mean thermal histories for gases, the radiant source temperature, and coal appear in Fig. 5.18 for both gasifiers. They were based on the exit gas temperatures calculated by Bockelie et al. (2003) for various test cases in Shell and GEPS gasifiers. The maximum values for both gas and radiation source were assumed to match the hottest exit gas temperatures. The approaches to the maximum values are consistent with those from CFD of the coal injectors in the T-fired furnace from Section 5.5.2, except that the time scale was extended to account for the greater pressures and coal loadings in the gasifiers, compared to those in a T-fired utility furnace. The thermal history for the GEPS reactor omits the preliminary stage for water evaporation, which extends the cool isothermal soaking period but does not

Simulations with detailed chemistry

165 2500

2500

2000

TWALL TWALL

1500

1500

TP 1000

TP

TGAS

1000 TGAS 500

500 Shell Conditions 0 0.0

Temperature, °C

Temperature, °C

2000

0.1

0.2

0.3 Time, s

0.4

0.5

0.6 0.0

GEPS Conditions 0.1

0.2

0.3

0.4

0.5

0.6

0

Time, s

Fig. 5.18 Estimated mean thermal histories for (solid curves) the mean size in the coal suspension; (dot-dashed curves) carrier gas; and (dashed curves) the radiant source for (left) Shell and (right) GEPS conditions. Reproduced from Niksa S. Simulating volatiles conversion in dense burning coal suspensions. Part 3. Extrapolations to entrained flow gasification conditions. Fuel 2019d;252:841–47 with permission from Elsevier.

affect the chemistry under consideration because the water completely vaporizes before any volatiles are released with and without the impact of water evaporation. The coal suspension heats at about 2000°C/s during this stage but remains below 300°C. Consequently, this stage is regarded as nonreactive, and simply adds a 200 ms lag to the onset of chemistry in the jet. After the lag, the gases are heated to their maximum temperatures in less than 200 ms at approximately 104°C/s primarily by radiation feedback from the macroscale coal flame. Both the coal and entrainment gas heat at similar rates in the Shell gasifier, while gas temperatures are as much as 500°C cooler than coal temperatures because only the particles absorb the intense thermal radiation. Coal heating rates are slightly slower than gas heating rates in the GEPS gasifier so that gases become hotter by the end of the heating period. These coal thermal histories were used to evaluate rates and product distributions for primary devolatilization in the ChemNet simulations. The subject flowfields are premixed primary jets with no entrainment of syngas via recirculation, so their equivalent reactor networks are simply linear CSTR series with 40 reactors for Shell and 45 reactors for GEPS, where about half the CSTRs covered the initial, nonreactive period. The subbituminous and hv bituminous coals from the previous sections are used again, whereas lignite was substituted for the lv bituminous because low-rank coals are preferred in integrated coal gasification combined cycle (IGCC) coal gasification for their relatively fast gasification reactivities. The most variable operating conditions for both gasifiers are the O2/coal and steam/coal ratios, because these ratios are the means to operate the gasifiers to produce essentially uniform syngas compositions from a broad range of fuel quality (Table 5.3).

166

Process Chemistry of Coal Utilization

Table 5.3 Coal properties and gasifier inlet conditions.

PVM, dry Ash, dry C, daf H, daf O, daf N, daf S, daf

O2/coal H2O/coal FCOAL, kg/s FO2, kg/s FH2O, kg/s

Lignite

Subbituminous

hv bituminous

50.5 1.2 73.2 5.3 20.6 0.8 0.1

40.6 6.4 75.5 5.2 17.9 0.9 0.4

32.6 10.3 83.3 5.4 8.4 1.4 1.3

Shell

GEPS

Shell

GEPS

Shell

GEPS

0.73 0.02 3.17 2.32 0.06

0.84 0.54 3.17 2.67 1.71

0.77 0.02 2.63 2.02 0.05

0.88 0.53 2.63 2.31 1.38

0.90 0.07 2.39 2.15 0.17

0.95 0.64 2.39 2.27 1.53

Reported values for both gasifiers were compiled from several sources reported by Niksa (2019d) and interpolated to make the assignments in Table 5.3, along with atomic H/C and O/C ratios of the whole feedstreams. These latter ratios are evaluated as
 %Hdry + 2 steam H 12 18 coal ¼ C 1 %Cdry
 %Odry + O2 + 16 steam O 12 coal 18 coal ¼ %Cdry C 16

(5.26a)

(5.26b)

where the C-, H-, and O-contents are on a dry basis, and the coal feed rate is for dry coal, so that the steam/coal ratio also includes inherent coal moisture plus slurry moisture in GEPS cases. With Shell gasifiers, the O2/coal ratio is generally increased for coals of progressively greater rank to maintain H/C ratios from 0.80 to 1.00 and O/C ratios from 0.90 to 1.15. GEPS gasifiers are fed with coal slurry that has 65 wt.% coal, so steam/coal ratios fall between 0.50 and 0.70. The O2/coal ratios are generally increased for coals of progressively higher rank although, as for Shell gasifiers, the variation in this parameter for coals with C-contents over 77% is minimal. For all but the lowest rank coals, the H/C varied from 1.85 to 2.10, and O/C varied only from 1.54 to 1.65. The adjusted coal flow rates were 3.17 kg/s with lignite; 2.63 with subbituminous; and 2.39 with hv bituminous. The same grind was used with each coal and was specified with Rosin-Rammler PSDs to give a mean size of 44.4 μm for Shell conditions. The GEPS gasifier fires a grind much coarser and broader than the standard utility grind with a mean size

Simulations with detailed chemistry

167

of 205 μm, as needed to stabilize the slurry viscosity. Reactor pressures were the same with all coals at 2.3 MPa for Shell and 4.1 MPa for GEPS. The flame structure in Fig. 5.19 for Shell conditions with subbituminous conveys two distinct conversion time scales. During the first partial oxidation stage from 225 to 275 ms, most of the GHCs and char consume all the available O2 to generate mostly CO with small amounts of CO2 and H2O from homogeneous oxidation. Soot burnout was negligible. Due to the fast oxidation reactivity of this char and its fine grind, just under 90% of the char burns out upon ignition. During the second gas reforming stage from 275 to 350 ms, two additional C2 GHCs—C2H4 and C2H6—appear as intermediate species, while C2H2 and CH4 pass through their maximum concentrations before they are eliminated by 400 ms. Carbon dioxide, CO, H2, and H2O change to reflect the elimination of GHCs and water-gas shifting. Whereas the predicted effluent levels of CO and H2 in Fig. 5.19 are lower than ultimate syngas concentrations, these values are expected to relax toward accurate values upon addition of CO and H2 by char gasification, and the associated water-gas shifting throughout the gas quench cycle. The flame structure with the same coal under GEPS conditions also appears in Fig. 5.19, where the partial oxidation stage extends from 250 to 290 ms, and the reforming stage continues through 375 ms. The extent of char burnout is much lower than for Shell conditions due to the much coarser grind and, to a lesser extent, the lower ultimate temperatures for gases and the radiant source, and the slightly lower O2 partial pressures for GEPS. Soot oxidation was again negligible. Since char oxidation is at a disadvantage with volatiles in the O2 competition, it releases relatively little CO. Consequently, the gas phase remains relatively oxidizing as it generates twice as much CO2 as Shell conditions. This more oxidizing environment also significantly alters the course of the reforming stage. Methane, C2H2, and C2H4 appear early in this stage, albeit at less than one-third of the levels under Shell conditions. But all GHCs are partially oxidized into CO and H2 by 385 ms. Consequently, the effluent from the near-burner flame zone contains only the four major syngas components. It is interesting that neither volatiles combustion nor a major portion of char oxidation perturbs the moisture level from its initial value, despite major shifts in the other syngas concentrations. Of course, both the CO and H2 concentrations will more-thandouble after a large amount of residual char is gasified by steam during the remaining residence time, and the gas composition is shifted during the quench cycle. Since char oxidation determines the gas composition leaving the primary jet, it is worth resolving the burnout across the PSD at the exit of the near-injector zone. These size dependences appear in Fig. 5.20 with the cumulative PSDs of the dry and slurry coal feeds. The smallest 60% of the GEPS PSD covers the entire Shell PSD and the full PSD extends to a millimeter, although the calculations do not account for fragmentation by thermal shock, internal pressurization, or any other means that may reduce the top size. The relations between burnout and size are similar for both operating conditions for all but the smallest sizes in the Shell PSD and the largest sizes in the GEPS PSD. Burnout is slightly lower for GEPS because the maximum reactor temperatures are slightly cooler (cf. Fig. 5.18), although the maximum char temperatures in the ChemNet simulations for both cases approached 2700°C. Char oxidation is inhibited for the largest GEPS sizes by their large thermal

168

Process Chemistry of Coal Utilization

100

80 Char

60

60 O2

40

20

40

subbituminous Shell Conditions

subbituminous GEPS Conditions

0

0 1.0 subbituminous GEPS Conditions

subbituminous Shell Conditions

Concentration, %

0.8

CH4

2.0 C2H2

1.5

0.6

CH4 C2H2

0.4

1.0 C2H4

0.5

0.2

C2H4

C2H6

0.0 70

0.0 subbituminous Shell Conditions

CO

70

subbituminous GEPS Conditions

60

60 Concentration, %

Concentration, %

2.5

20

H2O

50

50 40

40 CO

30

30

H2

20

20

H2O

CO2

10

Concentration, %

Burnout or Concentration, %

Char

80

Burnout or Concentration, %

100 O2

H2

CO2

10 0

0 200

250

300 Time, ms

350

400

200

250

300

350 Time, ms

400

450

Fig. 5.19 In descending order, predicted flame structures for (left) Shell and (right) GEPS conditions with a subbituminous coal including (top) (solid curve) extents of char burnout and (dashed curve) O2 concentration; (middle) the concentrations of (solid curve) CH4, (dashed curve) C2H2, (dotted curve) C2H4, and (lower solid curve) C2H6; and (bottom) concentrations of (dot-dashed curve) H2, (solid curve) CO, (dotted curve) CO2, and (dashed curve) H2O (Niksa, 2019c). Note change of scales between Shell and GEPS conditions on the time axes and on molar concentration in the middle panels. Reproduced from Niksa S. Simulating volatiles conversion in dense burning coal suspensions. Part 3. Extrapolations to entrained flow gasification conditions. Fuel 2019d;252:841–47 with permission from Elsevier.

Simulations with detailed chemistry

Char BO & Cum PSD, %

100

169

Shell BO

80

GEPS PSD Shell PSD

60

40 GEPS BO 20 subbituminous 0

0

50

100

150 200

500

750 1000 1250 1500

Particle Size, mm

Fig. 5.20 (Dashed curves) Cumulative Rosin-Rammler PSDs for the Shell and GEPS coal grinds; and predicted char burnout vs initial coal size for subbituminous coal at the exit of the NBFZ under (solid curve) Shell and (dot-dashed curve) GEPS conditions. Reproduced from Niksa S. Simulating volatiles conversion in dense burning coal suspensions. Part 3. Extrapolations to entrained flow gasification conditions. Fuel 2019d;252:841–47 with permission from Elsevier.

capacitances, which prevent these particles from achieving a fully ignited state. As coal size diminished from 2640 to 423 μm, the maximum char temperatures increased from 261°C to 1921°C, which increased burnout from 0% to 25%. Due to their relatively fast heat loss rates, the smallest Shell particles ignited but only to a slow-burn state that sustained cool temperatures. Maximum char temperatures were only 591°C and 646°C for sizes of 4 and 11 μm, respectively, which kept extents of burnout to only 12% and 27%. The GEPS PSD does not contain any particles this small, and the predicted burnout for its minimum size of 16 μm exceeds 98% under both Shell and GEPS operating conditions. The major gas components with lignite and hv bituminous under Shell conditions appear in Fig. 5.21. The flame structure with lignite is very similar to that with the subbituminous: the extent of char burnout is only slightly greater; all GHC concentrations are 25% lower; the nascent syngas contains slightly less CO and H2 and slightly more CO2 and H2O. In contrast, the hv bituminous conversion is only slightly greater than 60%, whereas the ultimate GHC concentrations are nearly identical to the subbituminous coal’s. The lower extent of char burnout is responsible for the substantially lower concentrations of CO and H2 in Fig. 5.21, and the greater CO2 and H2O levels. As more residual char leaves the NBFZ with the hv bituminous, the CO and

170

Process Chemistry of Coal Utilization

70

hv bituminous Shell Conditions

lignite Shell Conditions

CO

70 60

CO

50

50

40

40

30 20

30

H2O

H2 H2O

10

CO2

20

H2

10

Concentration, %

Concentration, %

60

CO2 0

0 200

250

300 Time, ms

350

400

200

250

300 Time, ms

350

400

Fig. 5.21 Predicted concentrations of (dot-dashed curves) H2, (solid curves) CO, (dotted curves) CO2, and (dashed curves) H2O for Shell conditions with (left) lignite and (right) hv bituminous. Reproduced from Niksa S. Simulating volatiles conversion in dense burning coal suspensions. Part 3. Extrapolations to entrained flow gasification conditions. Fuel 2019d;252:841–47 with permission from Elsevier.

H2 concentrations in the exit syngas will be much greater than those in Fig. 5.21 once nearly all char has been gasified. Under GEPS conditions, the coal quality impacts are weaker than under Shell conditions because the differences among extents of char burnout are smaller. With lignite, burnout is only slightly greater than with subbituminous; GHC levels are 25% lower, as for Shell conditions; and ultimate CO2 concentrations are slightly greater, whereas CO and H2 levels are slightly lower. Gas concentrations with hv bituminous in Fig. 5.22 show more significant differences. The maximum GHC levels are onethird lower. Char burnout is 25% lower than with subbituminous, and the H2O concentration is 25% greater. There is also double the CO2 in the nascent syngas, but only half-as-much CO and H2 as with subbituminous. As for Shell conditions, the coal quality impacts on nascent syngas compositions from GEPS NBFZs are determined by char oxidation reactivities, in that progressively faster reactivities produce more reducing gas conditions. Soot oxidation with all coals under both sets of operating conditions was negligible, and none of the GHCs from any coal persisted out of either of these NBFZs. To summarize, disparate extents of char burnout are the distinguishing factor that interprets the much different gas compositions from the NBFZs in Shell and GEPS gasifiers, provided that the char reactivity is sufficient to effectively compete for the available O2 with GHCs and other volatile fuel components; i.e., with all but coals of the highest rank. As O2 pressures and maximum reactor temperatures are comparable, the coal grind is the only disparate operating condition affecting char burnout in these NBFZs. Moisture levels are also much greater in GEPS gasifiers, but moisture acts much more as a diluent than as a reactant in primary flame zones. Consequently, the much finer coal grind in Shell gasifiers competes much more effectively for O2 with GHCs and other volatile fuel components. All but the smallest sizes reached

Simulations with detailed chemistry

171

1.0 hv bituminous GEPS Conditions

Concentration, %

0.8

0.6

0.4

CH4 C2H2

0.2

C2H4

0.0 70

hv bituminous GEPS Conditions

H2O

Concentration, %

60 50 40 30

CO2

20

CO

10 H2 0 200

250

300

350

400

450

Time, ms

Fig. 5.22 Predicted concentrations of (top) (solid curve) CH4, (dashed curve) C2H2, (dotted curve) C2H4, and (lower solid curve) C2H6; and (bottom) (dot-dashed curve) H2, (solid curve) CO, (dotted curve) CO2, and (dashed curve) H2O for GEPS conditions with hv bituminous. Reproduced from Niksa S. Simulating volatiles conversion in dense burning coal suspensions. Part 3. Extrapolations to entrained flow gasification conditions. Fuel 2019d;252:841–47 with permission from Elsevier.

or exceeded 90% burnout with low-rank coals under Shell conditions; but the largest quarter of the GEPS PSD could not achieve a fully ignited, rapid-burn state under comparable conditions, because of its inordinate thermal capacitance. This feature explains most of the difference between extents of burnout in both gasifiers with the same coal type.

172

Process Chemistry of Coal Utilization

The interesting qualification is that the smallest sizes in the Shell PSD also failed to fully ignite to a rapid-burn condition, due to their relatively rapid heat loss rates. The two smallest size increments of 11 and 4 μm are actually major sources of residual char from the flame zone because, in combination, they account for almost 10% of the coal grind in these simulations. Of course, the quantitative impact should not be generalized because different operators process different grinds. But as it is practically impossible to produce coal grinds whose mean sizes are smaller than those in p. f. firing without generating appreciable portions under 10 μm, disproportionate amounts of fine-particle residual char from primary flame zones should be evaluated as potential sources of unreacted carbon from Shell gasifiers, particularly when it travels in quench layers along cooler refractory walls. The other determining factor on char burnout is the extremely hot maximum temperatures of fully ignited char particles. Due to the combination of very hot maximum reactor temperatures and very high O2 partial pressures, chars of all three coal types were heated in excess of 2500°C during their burnout histories under both sets of operating conditions. Such temperatures are hot enough to completely deactivate the char reactivity, which is why char burnout is confined to such short time intervals in the dynamic flame structures. Once chars are exposed to such hot temperatures, they spontaneously extinguish before all the available O2 has been consumed due to the drastic reductions in reactivity from thermal annealing. Moreover, thermal annealing during char burnout also diminishes the gasification reactivities toward steam and CO2 beyond primary flame zones, which ultimately determine UBC levels in the ultimate syngas. Char burnout affects syngas compositions from primary flame zones by making the gas atmosphere more reducing for progressively greater extents of burnout. This tendency is apparent in greater levels of CO and H2 and lesser amounts of CO2 and H2O for progressively greater burnout. It explains the different ultimate compositions of nascent syngas from Shell and GEPS primary flames with the same coals, and also explains the variations for coals of progressively higher rank in either gasifier flame zone. The simulations also showed that GHCs are eliminated from flame effluents under both Shell and GEPS conditions with all coals. This circumscribes the finding that volatiles combustion under commercial loadings in p. f. firing alters the species concentrations but does not necessarily add or eliminate any components of secondary volatiles. Effluents from gasifier flame zones contain only the four major syngas components, along with minor levels of N- and S-species. Notwithstanding, the estimated gas compositions in these simulations are incompatible with the instantaneous combustion and equilibrium gas compositions within primary flames that are usually imposed in CFD gasifier simulations, because they are strongly influenced by simultaneous and very large extents of char burnout.

5.6

Summary

Regions near the fuel injectors or burners in pyrolyzers, furnaces, and gasifiers stage most of the chemistry of coal conversion, including drying; primary devolatilization; tar decomposition; secondary volatiles pyrolysis, reforming, and combustion;

Simulations with detailed chemistry

173

the combustion of char and soot; and, perhaps, gasification of char and soot. Under commercial fuel injection conditions, these regions are traversed in only a few hundred milliseconds of transit time. But this does not ensure that any of the chemical processes will be absent, because primary coal suspensions at commercial coal loadings heat to 1000°C or hotter in about 100 ms. Moreover, the heat release from volatiles combustion and char ignition drives suspension temperatures above the threshold for diffusionlimited burnout in additional tens of milliseconds. So the available transit times are long enough to finish primary devolatilization and tar decomposition, oxidize primary volatiles, and consume major portions of char and some soot via oxidation. Beyond these primary coal jets, residual gaseous fuel compounds, soot, and char will burn out if oxidizing streams are entrained, or be reformed and gasified in gasifiers. These scenarios are extremely demanding of simulation strategies. CFD simulations entirely omit the chemistry of volatiles conversion by using only the compositions of volatiles combustion products at equilibrium and setting the overall burning rates equal to turbulent mixing rates of secondary air streams into primary coal jets. One problem with this approach is that it does not distinguish between primary and secondary products, so soot is omitted. Soot is often the slowest burning fuel component in the system and sequesters major portions of what began as GHCs, which have the fastest burning rates of all fuel components. An even larger problem is that the equilibrium analysis for volatiles combustion also explicitly decouples the conversion of char from the combustion of the gaseous fuels. In actuality, all the fuel components—gaseous fuels, soot, and char—compete for the O2 in primary coal suspensions. In isolation, gaseous fuels burn faster than chars which, when derived from low rank and bituminous coals, burn faster than soot. But under the heavy coal loadings in primary fuel jets, all these fuels undergo simultaneous decomposition, reforming, and combustion. The only legitimate way to analyze this competition is with robust, validated reaction kinetics for every chemical process. And equivalent reactor networks are the only means yet devised to incorporate truly comprehensive kinetics for all fuel components. Indeed, the species conservation laws presented in this chapter explicitly show how the various chemical processes interact, and how variations in the operating conditions affect the competition. ChemNet postprocessing uses some of the information in a conventional CFD simulation to develop an equivalent reactor network for the flowfields in suspension-fired coal utilization technologies. The analysis largely omits the species concentration fields, because they are undermined by the rudimentary CFD chemistry submodels. But the flowfields and temperature fields are analyzed to specify the network configuration and the operating conditions for every reactor in the network. Networks developed this way are equivalent to the CFD results in the following ways: The RTDs in the major flow regions are the same in the CFD flowfield and in the section of the reactor network that represents the flow region under consideration. Mean gas temperature histories and the effective ambient temperature for radiant heat transfer are also the same. The entrainment rates of surrounding fluid into a particular flow region are evaluated directly from the CFD simulation. To the extent that the RTD, thermal history, and entrainment rates are similar in the CFD flowfield and reactor network, the chemical kinetics evaluated in the network represents the chemistry in

174

Process Chemistry of Coal Utilization

the CFD flowfield, albeit only in a steady, spatially averaged way. While the analysis supersedes the rudimentary chemistry in CFD, it completely omits the impact of turbulent fluctuations on the chemistry and all other aspects of the flowfield. ChemNet’s main conceptual advantages are that fully validated elementary chemical reaction mechanisms are implemented without approximation to describe volatiles conversion, NOX production and abatement, and residual CO levels in flue gas and syngas. In combination with comprehensive reaction mechanisms for devolatilization and char conversion, the analysis takes full advantage of the phenomenal advances in coal conversion chemistry over the past few decades. Moreover, realistic, comprehensive reaction mechanisms have many fewer adjustable parameters than rudimentary global reaction schemes. Collectively, the global rates in a conventional CFD chemistry submodel contain a few dozen kinetic parameters that must be adjusted for every coal. In contrast, the suite of comprehensive mechanisms in this section contains just three: initial char reactivities for oxidation and gasification, plus the fraction of char-N converted to NO during char burnout. Consequently, the amount of calibration data required to depict the distinctive behavior of a particular coal sample is far smaller with the comprehensive mechanisms. And the only sample-specific information they require are proximate and ultimate analyses for every coal sample. Now the presentation of ChemNet moves from the conceptual plane to the practical applications in Chapters 6 and 7.

References Antifora A, Faravelli T, Kandamby N, Ranzi E, Sala M, Vigevano L. Comparison between two complementary approaches for predicting NOX emissions in the furnaces of utility boilers. In: Fifth international conference on technologies and combustion technologies for a clean environment. Lisbon, Portugal: Calouste Gulbenkian Foundation; 1999. p. 377–84. Benedetto D, Pasini S, Ranzi E, Faravelli T, La Marca C, Tognotti L. NOX reduction evaluation in low-NOX combustion systems using 3D simulation and reactor network analysis. In: Fourth international conference on technologies and combustion technologies for a clean environment. Lisbon, Portugal: Calouste Gulbenkian Foundation; 1997. Benedetto D, Pasini S, Falcitelli C, La Marca C, Tognotti L. NOX emission prediction from 3-D complete modeling to reactor network analysis. Combust Sci Technol 2000;153:279–94. Bockelie MJ, Denison MK, Chen Z, Senior CL, Sarofim AF. Using models to select operating conditions for gasifiers. In: Proc. Pittsburgh coal conf., Pittsburgh, PA; 2003. Cao Z, Li T, Zhang Q, Zhou H, Song C, Fengqi Y. Systems modeling, simulation, and analysis for robust operations and improved design of entrained-flow pulverized coal gasifiers. Energy 2018;148:941–64. Curran HJ, Gaffuri P, Pitz WJ, Westbrook CK. A comprehensive modeling study of n-heptane oxidation. Combust Flame 1998;114:149–77. Ehrhardt K, Toqan M, Jansohn P, Teare JD, Beer JM, Sybon G, Leuckel W. Modeling of NOX reburning in a pilot scale furnace using detailed reaction kinetics. Combust Sci Technol 1998;131:131–46. Fogler HS. Elements of chemical reaction engineering. 2nd ed. Upper Saddle River, NJ: Prentice Hall PTR; 1993.

Simulations with detailed chemistry

175

Glarborg P, Alzueta MU, Dam-Johansen K, Miller JA. Kinetic modeling of hydrocarbon/nitric oxide interactions in a flow reactor. Combust Flame 1998;115:1–27. Liu G-S, Niksa S. Coal conversion submodels for design applications at elevated pressures. Part II. Char gasification. Prog Energy Combust Sci 2004;30(6):697–717. Luo Y, Wen X, Wang H, Luo K, Jin H, Fan J. An a priori study of different tabulation methods for turbulent pulverised coal combustion. Combust Theor Model 2018;22. Luo K, Zhao C, Wen X, Gao Z, Bai Y, Xing J, Fan J. A priori study of an extended flamelet/ progress variable model for NO prediction in pulverized coal flames. Energy 2019;175: 768–80. McConnell J, Sutherland JC. Assessment of various tar and soot methods and a priori analysis of the flamelet model for use in coal combustion simulation. Fuel 2020;265:116775. Neoh KG, Howard JB, Sarofim AE. Soot oxidation in flames. In: Siegla DC, Smith GW, editors. Particulate carbon formation during combustion. NY: Plenum Press; 1981. p. 261–82. Niksa S. FLASHCHAIN® theory for rapid coal devolatilization kinetics. 9. Decomposition mechanism for tars from various coals. Energy Fuel 2017;31:9080–93. Niksa S. Simulating volatiles conversion in dense burning coal suspensions. Part 1. Validation of reaction mechanisms. Fuel 2019a;252:821–31. Niksa S. Predicting ultimate soot yields from any coal. Proc Combust Inst 2019b;37:2757–64. Niksa S. Simulating volatiles conversion in dense burning coal suspensions. Part 2. Extrapolations to commercial p. f. firing conditions. Fuel 2019c;252:832–40. Niksa S. Simulating volatiles conversion in dense burning coal suspensions. Part 3. Extrapolations to entrained flow gasification conditions. Fuel 2019d;252:841–7. Niksa S, Liu G-S, Hurt RH. Coal conversion submodels for design applications at elevated pressures. Part I. Devolatilization and char oxidation. Prog Energy Combust Sci 2003;29:425–77. Niksa S, Sakurai Y, Fujiwara N. Predicting the conversion efficiencies of any coal type in CFBCs. Energy Fuel 2017;31:4507–19. Park C, Appleton JP. Shock-tube measurements of soot oxidation rates. Combust Flame 1973;20:369–79. Pedersen LS, Glarborg P, Dam-Johansen K, Hepburn P, Hesselmann G. A chemical engineering model for predicting NO emissions and burnout from pulverized coal flames. Combust Sci Technol 1998a;132:251–314. Pedersen LS, Glarborg P, Dam-Johansen K. A reduced reaction scheme for volatile nitrogen conversion in coal combustion. Combust Sci Technol 1998b;131:193–223. Pitz WJ, Seiser R, Bozzelli JW, Seshadri K, Chen C-J, Da Costa I, Fournet R, Billaud F, BattinLeclerc F, Westbrook CK. Chemical kinetic study of toluene oxidation under premixed and non-premixed conditions [UCRL-CONF-201575]; 2003. Reith M, Proch F, Clements AG, Rabacal M, Kempf AM. Highly resolved flamelet LES of a semi-industrial scale coal furnace. Proc Combust Inst 2017;36:3371–9. Roessler DM, Faxvog FR, Stevenson R, Smith GW. Optical properties and morphology of particulate carbon: variation with air/fuel ratio. In: Siegla DC, Smith GW, editors. Particulate carbon formation during combustion. NY: Plenum Press; 1981. p. 57–90. Seiser H, Pitsch H, Seshadri K, Pitz WJ, Curran HJ. Extinction and autoignition of n-heptane in counterflow configuration. Proc Combust Inst 2000;28:2029–37. Shinnar R. Use of residence- and contact-time distributions in reactor design. In: Carberry JJ, Varma A, editors. Chemical reaction and reactor engineering. New York, NY: Marcel Dekker, Inc.; 1987. Van der Lans RP, Glarborg P, Dam-Johansen K, Knudsen P, Hesselmann G, Hepburn P. Influence of coal quality on combustion performance. Fuel 1998;77(12):1317–28.

176

Process Chemistry of Coal Utilization

Watanabe J, Okazaki T, Yamamoto K, Kuramashi K, Baba A. Large-eddy simulation of pulverized coal combustion using flamelet model. Proc Combust Inst 2017;36:2155–63. Wen X, Luo K, Luo Y, Kassem HI, Jin H, Fan J. Large eddy simulation of a semi-industrial scale coal furnace using non-adiabatic three-stream flamelet progress variable model. Appl Energy 2016;183:1086–97. Wen X, Rieth M, Scholtissek A, Stein OT, Wang H, Luo K, Kempf AM, Kronenburg A, Fan J, Hasse C. A comprehensive study of flamelet tabulation methods for pulverized coal combustion in a turbulent mixing layer—part I: a priori and budget analyses. Combust Flame 2020a;216:439–52. Wen X, Rieth M, Scholtissek A, Stein OT, Wang H, Luo K, Kronenburg A, Fan J, Hasse C. A comprehensive study of flamelet tabulation methods for pulverized coal combustion in a turbulent mixing layer—part II: strong heat losses and multi-mode combustion. Combust Flame 2020b;216:453–67. Zhao C, Luo K, Cai R, Xing J, Gao Z, Fan J. Large eddy simulations and analysis of NO emission characteristics in a laboratory pulverized coal flame. Fuel 2020;279:118316.

ChemNet furnace applications

6

Nomenclature af F0C FC, C FC, EF FjE FjP F∞ S jD M C0 Mi MS 0 N V x XjC XSj yjE, i Yi yjP, i yPAHi

radius of char fines, cm mass flowrate of char based on its ultimate yield, g/s carryover rate of char from a dense bottom bed, g/s ejection rate of char from a dense bottom bed, g/s gas entrainment increment into CSTR j primary gas stream into CSTR j mass flowrate of soot based on its ultimate yield, g/s number of spatial dimensions j in a simulation char molecular weight based on its elemental composition, g/mol molecular weight of species i soot molecular weight based on its elemental composition, g/mol number of CSTRs in a series for a flow region volume of a CSTR hypothetical fraction of PAH remaining after soot production extent of char burnout in CSTR j extent of soot burnout in CSTR j mass fraction of species i in gas entrainment increment into CSTR j species yield in secondary pyrolysis products, daf wt% mass fraction of species i in primary gas stream into CSTR j mass fractions of three surrogate PAH compounds to analyze VOC emissions

Greek symbols η vC,i,O2 νS,i,O2 ωi τ

effectiveness factor for O2 penetration into a burning char particle stoichiometry for species i in char oxidation, mol/mol of combustibles in char stoichiometry for species i in soot oxidation, mol/mol of combustibles in soot net species production rate from homogeneous chemistry, moles/cm3-s residence time within a flow region

Subscripts C E i O2 P S

char entrainment increment into a CSTR molecular species i pertaining to oxidation primary gas stream into a CSTR soot

Process Chemistry of Coal Utilization. https://doi.org/10.1016/B978-0-323-89959-8.00006-9 Copyright © 2022 Elsevier Ltd. All rights reserved.

178

Process Chemistry of Coal Utilization

This chapter presents numerous case studies based on ChemNet CFD post-processing to characterize in detail the chemical structures and emissions for pc and circulating fluidized-bed combustion (CFBC) furnaces. It contains a few cases for lab-scale systems, but the primary emphasis is on systems at pilot- and commercial-scale. The cases cover coal conversion and emissions for pc furnaces at lab-, pilot-, and commercial-scale and a series of commercial-scale CFBCs. They demonstrate that ChemNet delivers predictions for pollutants and unconverted fuel species that are appreciably more accurate than CFD results. At face value, this goal can be met in straightforward comparisons between measured and predicted values for the output variables of interest. The next section discusses the multitude of factors that can undermine “straightforward comparisons” with data from most pilot-scale systems and all commercial-scale furnaces. Then case studies are presented in sequence for pc firing configurations at a progressively larger scale and CFBCs.

6.1

ChemNet validation with data from large-scale systems

Measurement uncertainties inevitably increase for furnaces at a progressively larger scale. Under the best circumstances, a lab-scale furnace is supported by known uncertainties on all inputs and outputs, and diagnostics are applied throughout all intermediate stages of the coal conversion process. Detailed characterizations from start to finish are especially stringent in the validations of ChemNet simulations because they expose situations where the simulations give accurate output predictions but for incorrect reaction dynamics. The fluid mechanics in lab tests can often be simplified to display the stages of conversion over a single spatial dimension that has an explicit equivalence to transit time. This time coordinate is the same as the time coordinate in a ChemNet simulation, so the comparison of measured and predicted reaction dynamics is straightforward. Nonetheless, even conventional lab-scale systems can be subject to inordinate uncertainties on essential operating conditions. Consider the thermal history for fuel particles in tests that recreate suspension firing conditions. Flat-flame burners are subject to the least uncertainty, because individual coal particles are injected in a very weak carrier gas stream into a much larger isothermal flow of preheated gases. The carrier gas is rapidly heated and dispersed into the reactor flow, while individual particles are rapidly heated by convection from the isothermal flow with radiation transfer to the surroundings at a known temperature. Particle temperatures cannot be measured from the point of injection, but their thermal incandescence can be detected once the particle temperature approaches 800°C to 1000°C, depending on the size. These measurements often agree with transient particle temperatures calculated from simple enthalpy balances where the main source of uncertainty is on the thermophysical properties of coal, which are manageable. Drop tube tests are similar, in that very dilute particle streams fall into a slow isothermal flow within a heated tube. The flow tubes are opaque which makes it much harder to monitor temperatures

ChemNet furnace applications

179

of individual particles, but the heat transfer to the particle is still the same as in flatflame burners and amenable to accurate calculation. But when coal suspension loadings in laminar or turbulent entrained-flow reactors are increased to even minor fractions of the loadings in commercial furnaces and gasifiers, the uncertainties surge. The mixing and dispersion of a dense coal stream into the preheated gas flow surrounding a fuel injector substantially complicates the thermal analysis. If the coal and gas flows are matched, then the thermal field is necessarily two-dimensional downstream of the fuel injector; and if the mixing is intensified by swirl, then the analysis is subject to the substantial uncertainties of turbulent mixing of two-phase flows. These uncertainties are particularly important in coal processing because devolatilization occurs while the coal is still heating to the reactor temperature at even the fastest heating rates. The best that can be done is to assess temperature gradients with axisymmetric CFD of the actual fuel injector components and inlet conditions, and evaluate how these uncertainties propagate into any parameter assignments, such as reaction rate constants. At the pilot-scale, the uncertainties on all input and output variables are still wellknown, but flow fields are never one-dimensional. Moreover, imperatives for flame stability usually require swirled secondary and tertiary air streams, which move the flowfields into three dimensions, at least near the fuel injectors or burners. As long as the flows remain axisymmetric along the length of a unit, they can be diagnosed with radial sampling traverses. But there are often substantial asymmetries that operators may not even be aware of. In the extreme, systems may display different operating modes on different test days due to instabilities in the flows of some essential feed stream, or to some imprecise mechanical setting, or unregulated vibration potential, or any number of obscure factors. Under the worst circumstances, datasets have to be segregated by test day, with some data groups attributed to normal operations and others regarded as flawed, albeit for unknown reasons. Nothing is straightforward in any evaluation that involves field testing at any commercial-scale coal utilization technology for numerous reasons. As explained at the end of Section 2.1, even the standard coal properties are ambiguous at any full-scale pc furnace. There are no fuel injection conditions per se; only ambiguous mechanical settings on portions of multiport fuel delivery manifolds that are supposed to “balance” the distribution of coal to the various burners or fuel injectors. Serious fuel maldistribution may be apparent as fuel impingement on a furnace waterwall section, or in near-burner slagging patterns. Or it may not be recognized at all. There are no diagnostics for the burner belts in commercial furnaces except for infrared (IR) emission detectors that monitor what is often called a temperature but has much less meaning than that. Flows within commercial pc furnaces are never homogeneous for many reasons. Flows from burners or injectors often partition into central core flows that carry the bulk of the coal and mostly gas streams that move along furnace waterwalls. Overfire air streams are injected within burner belts as well as at much higher elevations. These different streams do not mix into a homogeneous flue gas flow by the furnace exit, as evident from significant transverse gradients in pollutant levels and, to a lesser extent, flyash loadings. The point of this litany of uncertainty is to emphasize that the potential to validate simulation results diminishes and, ultimately, vanishes in the progression from lab- to

180

Process Chemistry of Coal Utilization

pilot- to commercial-scale coal-fired furnaces. Not even all lab-scale systems provide data that is suitably stringent for validation work, because accurate thermal histories for every coal stream are an absolute requirement. But some lab-scale systems do achieve stringent validations for even the most complex chemical simulations, as exemplified by the case studies in this chapter. With pilot-scale systems, the stringency of any validation dataset falls off with distance from the furnace exit toward the fuel injector or burner. In near-burner flame zones (NBFZs), measured temperatures may be subject to tolerable measurement uncertainties, but it is difficult to decide whether they can be attributed to either the gas or coal phase rather than both. And measured levels of O2 and pollutants are merely suggestive because no suite of such measurements has yet closed material and element balances with the inlet stream conditions. Further downstream these balances can be closed once gas compositions have diminished to mixtures of mostly steam, CO2, CO, and H2 with unconverted soot and char combustibles. At the furnace exit, flue gas is often thoroughly mixed, so the measured levels of pollutants and other trace species are amenable to quantitative interpretations. Datasets from commercial-scale can only be taken at face value with respect to the uncertainties of the measurement devices. There can be no closures of mass and element balances because of the substantial uncertainties on the coal properties, as-fed; on the firing rate when the data were recorded; and on the uniformity of the streams being sampled at the furnace exit. For all these reasons, the quantitative validation of simulation results is confined to datasets from lab-scale systems. Given measurements or calculations that accurately specify the thermal histories of coal and carrier gas and suitable indices on the dynamic conversion of all fuel components and on the dynamic production of all species of interest, simulation results can be directly compared to measurements resolved in transit time. Such a comparison can establish whether the simulation results are within measurement uncertainties, provided that the suite of measurements close mass and element balances within tight tolerances, ideally within 5%. Otherwise, the agreement is merely suggestive. With datasets from pilot-scale systems, the same degree of validation can be established once the flowfield has relaxed its thermal and concentration gradients, and become nominally one-dimensional. Problem is, this stipulation excludes the entire near-burner region where most of the coal is converted and nearly all of the trace species are produced. The omission lets in an even deeper, conceptual problem: Gas compositions near the exit of a furnace or gasifier are integral in nature and can, therefore, be matched by an infinite number of trajectories through the excluded, upstream portion of the system. In other words, an infinite number of mathematical models can accurately match characteristics at the exit of a system even though their dynamics are wildly different. Nonetheless, one still can assert that agreement among measured and predicted exit characteristics are within measurement uncertainties, provided that the suite of measurements close the mass and element balances. But this agreement does not establish the validity of the reaction mechanisms in the simulation; instead, it only establishes that the mechanisms are viable candidates, pending more stringent validation, presumably, at lab-scale where the dynamics are accessible. Conversely, discrepancies with the data mean that the simulations did not sustain the evaluation, and the reaction mechanisms are suspect.

ChemNet furnace applications

181

This author describes agreement among data and simulations under these circumstances with the phrase, “within useful quantitative tolerances.” Presumably, simulations were developed in the first place to demonstrate that they can reproduce the measured exit characteristics within some externally set tolerance. For furnace CO emissions that tolerance may be quite large, such as within a factor of two, whereas for NOX emissions that may be within 15 ppm on a total concentration of several hundred ppm. Neither of these tolerances may have anything to do with the repeatability and precision of the actual detectors because the values are being used to satisfy external constraints, such as compliance with emissions regulations. If a manager decides that NOX emissions from a pilot-scale furnace must be predicted within 25 ppm for a suite of diverse coals to select among several candidate control strategies, then simulations that agree to furnace exit NOX data within this range are within useful quantitative tolerances. Moreover, that tolerance may be satisfied only for the mean value of several repeated measurements if the spread in the data exceeds the stated tolerance. Engineers can be fond of multiple qualifications on every aspect of their findings. But managers usually apply simple metrics without any qualifications at all to emphasize the practical utility over and above the statistical variance. This is precisely the distinction drawn by the “useful” in “within useful quantitative tolerances.” Reaction mechanisms should be validated at a lab-scale to demonstrate that they are able to accurately depict the reaction dynamics, preferably over the full domain of conditions in the commercial application. ChemNet simulations are the only means currently available to do this with truly comprehensive reaction mechanisms. Once the validation is sustained at lab-scale, ChemNet simulations with the same mechanisms can be validated with datasets from pilot-scale to contend with many of the complications that arise at commercial-scale, especially turbulent mixing effects. In the ideal situation, a CFD simulation of the pilot-scale flowfield carries all the essential aspects of mixing and entrainment into the equivalent reactor network in the ChemNet simulation. When the same coals are tested at the lab- and pilot-scale, there may be no adjustable parameters whatsoever in the validations with datasets from the pilot-scale. Otherwise, there can be up to three adjustable parameters for conventional combustor performance indices, as explained in Chapter 5. In either case, agreement with datasets from the pilot-scale can demonstrate that the simulations accurately predict characteristics at the system exit within useful quantitative tolerances, but it does not validate the mechanisms, per se. Once a set of mechanisms has performed as well as necessary at pilot-scale, it can be used for commercial-scale systems. But comparisons among measurements and simulation results have nothing to do with validations of reaction mechanisms or any other aspects of the simulations. For reasons already enumerated, any agreement between field test data and simulation results does not firmly establish anything about the simulations or the reaction mechanisms. These simulations are akin to virtual reality devices that generate sequences of stages in the chemical conversion of coal into flue gas. The process of developing an equivalent reactor network from the CFD simulation is often extremely informative, especially in the means to differentiate regions of the flowfield that sustain distinctive chemistry. The ultimate ChemNet simulations

182

Process Chemistry of Coal Utilization

apportion distinct aspects of coal conversion to specific regions of the flowfield and also indicate how variations in the operating conditions—thermal histories, RTDs, and entrainment histories—affect the conversion histories. When all is said and done, ChemNet simulations of commercial-scale systems give a much deeper understanding of when and where coal is converted and pollutants form; by which means these processes occur; and how the conditions can be regulated to improve the overall performance. But they really have very little to do with quantitatively validating any aspect of the simulations for commercial-scale.

6.2

Target variables for ChemNet simulations

Specifically, what are ChemNet simulations good for? That is, what species concentrations, conversion indices, etc. can be accurately determined with a ChemNet simulation? This section addresses this question in terms of the combustibles mass, heteroatomic speciation, and trace elements in coal. The combustibles in coal largely comprise its carbon, oxygen, and hydrogen and exclude its mineral matter and all minor and trace elements. In furnaces, the combustibles are burned to form steam and CO2. Chemical reaction mechanisms are superfluous to the concentrations of steam and CO2 in flue gas because these concentrations can be accurately estimated from a coal’s proximate and ultimate analyses and the excess O2 level in flue gas at a furnace exit. Unless the UBC in flyash is truly excessive, the flue gas composition is hardly perturbed by unconverted combustibles. Notwithstanding, UBC emissions from pc furnaces are legitimate targets for ChemNet simulations. All the heterogeneous mechanisms in Chapter 5 are needed to predict UBC emissions, whereas the homogeneous reaction mechanism may be streamlined by eliminating the reaction subset for N-species transformations. CBK/ E is among only a handful of char conversion mechanisms that describe the various deactivation modes during the latter stages of burnout, which is essential for accurate UBC or LOI predictions. Unfortunately, UBC emissions are also affected by a multitude of other factors, as described in Section 3.3. Even with measured grind PSDs and calibrated reactivities for char conversion kinetics, LOI predictions are complicated by secondary flows such as quench layers along waterwalls that may preserve disproportionate amounts of UBC through a furnace exit, and also by wall impingement, poor fuel distribution, and unbalanced air injection. None of these features are easy to replicate in CFD simulations or to carry over to an equivalent reactor network. Consequently, ChemNet simulations have been able to deliver accurate coal conversions for lab-scale furnaces but not for pilot- or commercial-scale systems. The situation for CO emissions is comparable. Certainly, FLASHCHAIN® gives reasonable predictions for CO from primary devolatilization and tar decomposition; the homogeneous mechanism accurately predicts intermediate CO production during volatiles combustion; and CBK/E accurately predicts CO levels during char burnout. But to an even greater extent than for LOI emissions, CO levels at a furnace exit are largely determined by chemistry in quench layers along furnace waterwalls. In turn, the CO levels in quench layers along upper furnace elevations are affected by coal and

ChemNet furnace applications

183

air maldistributions and the mixing intensities in burner belts. Contributions from reactions that “freeze” CO concentrations at the high levels associated with the much hotter temperatures in a burner belt are especially significant. Quench layers and coal maldistribution are not easy to replicate in a CFD simulation or an equivalent reactor network. Consequently, ChemNet simulations tend to underestimate CO exit concentrations, and the disparities with measured values would be greatest in poorly tuned furnaces that give the greatest CO emissions. The situation for volatile organic compounds (VOCs) is similar to CO’s in that these compounds only form in quench layers along waterwalls. Moreover, the homogeneous reaction mechanism to analyze their production in furnaces is considerably larger than the one used for homogeneous C/H/O/N chemistry, due to the addition of numerous heavier GHCs and light oxygenated species and intermediates. A case study on VOC emissions from ChemNet simulations is presented in Section 6.3.3.2. The transformation of coal-N into HCN, NH3, N2, N2O, and NO during combustion is the basis for aerodynamic NOX abatement technology. In practice, the goal is to minimize NO levels at a furnace exit by applying low-NOX burners, staging OFA at multiple furnace elevations and, perhaps, selective noncatalytic NOX reduction (SNCR) or hydrocarbon reburning upstream of convective passes. Even for systems with selective catalytic reduction (SCR) units, the imperative is to minimize the concentration of NO at a furnace exit to reduce the size and operating costs of the SCR. ChemNet simulations are suitable for analyses of near-burner NO production, as well as NO reductions due to furnace staging schemes, SNCR, and reburning. They have already interpreted the impact of fuel quality on furnace NOX emissions for coals across the rank spectrum with much better accuracy than CFD. Yet only a single coal-N conversion parameter in ChemNet is fuel-specific—the fraction of char-N converted into NO—whereas CFD chemistry submodels require several adjustments for every coal. At this point, all the reaction mechanisms have been validated for accurate NOX predictions for a diverse assortment of coals, and many case studies in this chapter feature comparisons among measured and predicted N-speciation. The transformations of coal-S into SO2, SO3, COS, and CS2 during combustion are also well-suited for ChemNet simulations, albeit with qualifications. The greatest ambiguities in the chemistry pertain to the release of coal-S during devolatilization. In particular, in FLASHCHAIN® (Niksa, 2017), coal-S must be allocated into aliphatic and aromatic S-functional groups even though the necessary analytical methods are not widely available. Consequently, the devolatilization of coal-S cannot currently be accurately predicted from a coal’s proximate and ultimate analyses; specialized analytical support is required. This is not a major hindrance in furnace applications because nearly all coal-S will ultimately be present in flue gas as SO2 under all furnace operating conditions. In furnaces, the most important S-transformation is the conversion of a small portion of the SO2 into SO3, which may accelerate corrosion rates in air preheaters. The conversion begins in the convective passes through homogeneous chemistry; then continues through an economizer with heterogeneous chemistry on Fe2O3 in flyash; then finishes with heterogeneous chemistry on SCR catalysts, if present in the gas cleaning system (Krishnakumar and Niksa, 2010). The first chemical stage across

184

Process Chemistry of Coal Utilization

the convective passes is amenable to ChemNet simulations where the turbulent flue gas flow is staged as PFRs-in-series, and the homogeneous reaction mechanism for Sspecies transformations contains 134 reactions among 30 species and no adjustable parameters (Yilmaz et al., 2006). Fig. 6.1 shows an evaluation of ChemNet SO3 predictions with 14 tests at 13 commercial furnaces (Krishnakumar and Niksa, 2010). The lowest labels on the X-axis give the furnace label, for which furnace L-S1 has two field tests. From left to right, two tests monitored SO3 levels across an SCR; three were for only the SCR outlet; five were for only the SCR inlet; one was for an air preheater inlet; and three were for ESP inlets. Twelve tests were run with bituminous coals, and the remainder were with either subbituminous or a subbituminous/bituminous blend. The coal-S varied from 0.4 to 4.8 daf wt%. As seen in Fig. 6.1, the predicted SO3 levels from 13 commercial furnaces are accurate with only one exception for furnace S5. The two datasets from L-S1 had economizer O2 levels of 3.5% and 2.1%, and SO3 was lower by 29.3% for the lower O2 level, vs. a predicted reduction of 29.4%. Sulfur-species transformations under oxidizing conditions can be accurately predicted with the fully validated reaction mechanisms in the literature and the simplest of all reactor networks. They are not considered further in this chapter, pending further comparisons with field test data on commercial systems. Many of the minor species in coal are suitable targets for ChemNet simulations, particularly halogens and the alkali and alkaline earth metals (AAEMs). The main differences with other heteroatoms are that halogen transformations within furnaces tend to align with thermochemical equilibrium at flame temperatures, and their heterogeneous chemistry on char, soot, and minerals is usually inconsequential. Consequently, the speciation for Cl and Br can be accurately estimated as equilibrium compositions up to the furnace exit. Into and along the gas cleaning system, ChemNet simulations have been developed with suitable homogeneous reaction mechanisms and a network

Fig. 6.1 (Open bars) Measured and (hatched bars) predicted SO3 concentrations at various sampling locations on a wet basis corrected to 3% O2. Coal-S values (▲) in daf wt% are indicated on the right axis (Krishnakumar and Niksa, 2010).

ChemNet furnace applications

185

of PFRs-in-series for the turbulent flow through the ductwork and cleaning units. The author’s group assembled a homogeneous mechanism with 252 reactions to analyze Hg emissions (Niksa et al., 2010). The Cl-speciation is uninteresting because only HCl is present in the oxidized flue gas in coal-fired furnaces. But the Br-speciation shows the rapid production of HBr across the economizer and a slower accumulation of Br2 through the downstream units. These transitions, in turn, promote Hg0 oxidation rates and thereby enhance the collection of Hg species upstream of the smokestack. The situation is comparable with AAEMs, albeit with considerably more complex chemistry. AAEMs factor into the performance of so-called “Advanced Reburning” in which the chlorides or oxides of Na and K are co-injected with a hydrocarbon fuel in the convective passes to promote reburning chemistry that reduces NO levels. AAEMs also spontaneously condense into oxides and chlorides that ultimately form sulfates that plug and poison SCR catalysts and produce ammonium bisulfates that can condense into blue plumes from smokestacks. Homogeneous reaction mechanisms are available for these transformations (Glarborg and Marshall, 2005), but the release of AAEMs during devolatilization and char conversion is complicated by their coalescence with various molten mineral phases. These interactions need to be built into the mechanisms for coal conversion, and the interactions with molten minerals add another phase to the reactors in ChemNet simulations. While possible in principle, these extensions require substantial additional development because they introduce new physical processes into the current calculation sequence. Trace metal transformations in ChemNet should be restricted to only those metals that are released into gases within a furnace, which are Hg and, to lesser extents, Se, As, and B. The analysis of Hg transformations is the most advanced, by far. ChemNet simulations of Hg speciation have already been used to quantitatively interpret a few hundred field tests on Hg-speciation at commercial-scale utility furnaces (Niksa and Krishnakumar, 2012). Coal-Hg is entirely released in furnace flame zones, and all Hg is the elemental metal (Hg0) at furnace exits. Minerals do not scavenge Hg in furnaces. The chemical transformations begin in economizers and move through a series of stages that involve both homogeneous and heterogeneous mechanisms. Without Br injection, heterogeneous chemistry on UBC and carbon sorbents predominates with adsorbed Cl on UBC as the primary oxidizer. With Br, the contributions from homogeneous chemistry are much more important (Niksa et al., 2010). The ChemNet interpretations of measured Hg-speciation at full-scale gas cleaning systems again used only PFRs-in-series for the turbulent flow through the gas cleaning system, in conjunction with the 252-step homogeneous reaction mechanism for Hg/ Cl/Br chemistry and multi-step heterogeneous mechanisms for the Hg chemistry on UBC and carbon sorbents. This application was developed from the methodology described in Chapter 5, except that those mechanisms for coal conversion and hydrocarbon and N-species conversion in the gas phase were exchanged with specialized heterogeneous mechanisms for Hg/Cl/Br chemistry on UBC and sorbents and a homogeneous mechanism for combined halogen/Hg chemistry. Once the mechanisms were exchanged as needed, the specification of operating conditions across the network and the ChemNet calculation sequence were essentially the same as described in Chapter 5.

186

Process Chemistry of Coal Utilization

The same approach has also been extended to Se transformations along utility furnaces and cleaning systems (Krishnakumar et al., 2013), although field tests to validate predicted Se-speciation are scarce and subject to much greater uncertainties than for Hg-speciation. Gaps in the reaction mechanisms for Se are significant but not insurmountable. However, for As and, particularly, B even the phenomenology is ambiguous, and comprehensive reaction mechanisms for homogenous chemistry and various stages of heterogeneous chemistry are still being formulated. All three metals have incomplete and ambiguous vaporizations during the initial stages of combustion, and their volatility falls off considerably from Se to As to B. Arsenic poisons SCR catalysts which is managed, as needed, with upstream injections of kaolinite or lime. Consequently, most of the available laboratory characterization work on As emphasizes As capture on inherent or injected minerals, rather than the chemistry of As transformations. Even less is known about B transformations although, again, control methods are the prominent research topic which, for B, focuses on precipitation out of the wastewater from FGD scrubbers. In summary, ChemNet postprocessing has already been used to accurately interpret NOX emissions from pilot- and commercial-scale furnaces, as well as important minor species (SO3, halogens) and the most volatile trace metal emissions (Hg, Se) from commercial furnaces. It is considerably less accurate for UBC and CO emissions, because these species are determined by conditions that are not well represented by CFD simulations of large, coal-fired furnaces and, therefore, difficult to transmit into equivalent reactor networks. AAEM transformations are suitable ChemNet targets but have not yet been analyzed, pending the extensions for the new physical processes of sulfate aerosol generation. Otherwise, the necessary reaction mechanisms are available. Extensions for As and B transformations are primarily limited by a dearth of fundamental characterization work at lab-scale, which would ultimately form the foundation for suitable chemical reaction mechanisms.

6.3

Case studies on PC combustion

This section presents a series of case studies on coal conversion during pc firing. It moves through cases at lab-, pilot-, and commercial-scale and includes at least one case at every scale that formally specified the equivalent reactor network from a conventional CFD simulation. Topical cases based on rudimentary, approximate networks are also included. All cases have coal loadings at the burners or fuel injectors that either match or approach those of 0.40 to 0.50 kg-coal/kg-primary-air in commercial furnaces and use grinds within the range of commercial pc firing. Diverse assortments of coals were processed in the cases at all three scales so the coal quality impacts are apparent. The goal is to resolve coal flame structure; i. e., the dynamic changes in the concentrations of all species of interest in a single spatial coordinate or in transit time through the entire combustion zone. While conceptually straightforward, such dynamic resolution is extremely challenging to achieve at even lab-scale, as seen in the next section.

ChemNet furnace applications

187

6.3.1 PC combustion at lab-scale The most direct means to resolve coal flame structure is in a single spatial dimension. But can 1D coal flames actually be operated continuously? The answer is yes, as demonstrated with a Meeker burner fired with subbituminous and hv bituminous coals run with loadings of 0.15 and 0.25 g-coal/g-carrier gas (Altenkirch et al., 1979; Peck et al., 1984, 1991). The coal was entrained in O2/Ar mixtures with respective coal-based SR values of 0.6 and 0.5 for the low loading, and 0.38 and 0.33 for the high loading. The major intermediates and products were monitored with sampling probes along the flow coordinate. However, with the low coal loadings, O2 was eliminated within 20 ms with the subbituminous and in half that time with the hv bituminous. With the high loading, O2 was eliminated in 5 ms with both coals. Consequently, these tests could not resolve the dynamic competition for O2 among primary and secondary noncondensables, soot, and char in the primary flame zone, although they provided one of the most thorough characterizations in the literature of NO reduction along postflame zones. Would it be possible to replace the sampling probes with optical diagnostics to obtain better time resolution? No, because the high coal loadings of interest are essentially opaque to diagnostic signals, especially when the high soot loadings during the initial stages of coal combustion compound the opacity. Could the O2 concentration be diminished to extend the combustion time scales to enable better time resolution with probes? Yes, but this strategy is limited by the stability limits of the coal burner, which requires a certain intensity of heat release for stable operation. Tests at the high loading in the Meeker burner were already close to the stability limit. The only means demonstrated so far to resolve the stages of fuel conversion in coal flames is to use O2 depletion as a “quench” on chemistry along a primary flame zone. When coal at a uniform suspension loading is entrained by an oxidizing carrier gas into a 1D flow reactor, increasing the inlet O2 level from zero through the level for stoichiometric combustion progressively moves the reaction system from secondary volatiles pyrolysis, through oxidative pyrolysis, through conversion of each of the gaseous fuels from fastest through slowest, through soot oxidation, and through char oxidation. Of course, these stages may overlap. But the value in such a test series is that it resolves the conversion of all the major fuel components which can, in turn, be directly related to fuel-N transformations. However, these benefits can only be realized if this hypothetical flow reactor can actually operate with inlet O2 levels as low as none. This is problematic for any conventional burner design that uses heat feedback to stabilize the coal flame. But the Radiant Coal Flow Reactor (RCFR) overcomes this difficulty (Chen and Niksa, 1992; Niksa and Cho, 1996) and is the subject of the two case studies at lab-scale in this chapter.

6.3.1.1 1D coal flame structure at atmospheric pressure The RCFR is a variation on an EFR. It was designed to monitor burning entrained coal suspensions at realistic coal loadings and heating rates without two-phase mixing or any stability envelope at all. It heats entrained coal suspensions by thermal radiation from a black-body enclosure, not by a preheated gas stream. As the entrainment stream is

188

Process Chemistry of Coal Utilization

transparent to radiation and the suspension is optically thin at even high loadings due to its small dimensions, coal thermal histories can be based on the same radiant flux to all particles in the suspension. The furnace system also includes rapid quenching to resolve reaction dynamics on a 10 ms-time scale, aerodynamic classification to segregate aerosol, particulate, and gaseous products, and analyses for complete product distributions. Mass and element balances typically close to within 5% in individual tests. The flowfield based on CFD appears in Fig. 6.2. Coal is entrained in Ar with a specified amount of O2 onto the centerline of a heated cylindrical tube to form a premixed

Fig. 6.2 (Top) CFD-based regions for the RCFR and (bottom) assigned thermal histories for solids and gases from (dashed curves) subbituminous, (solid curves) hv bituminous, and (dotted curves) lv bituminous coals. The estimated temperature of the radiation source appears as TWALL. (Reproduced with permission from Niksa S. Simulating volatiles conversion in dense burning coal suspensions. Part 1. Validation of reaction mechanisms. Fuel 2019a;252:821–31. Elsevier.)

ChemNet furnace applications

189

suspension with 0.15 g-coal/g-carrier. The suspension is surrounded by an annular sheath flow with the same O2 level and inlet velocity to minimize particle dispersion off the centerline. As a suspension moves through the tube in laminar flow, coal particles heat at rates exceeding 104°C/s to the onset temperature for primary devolatilization and release their volatiles, which then decomposes further in the gas phase, mix with O2, and burn. Fig. 6.2 also shows the assigned thermal histories for coal (TP), entrainment gas (TGAS), and the effective radiation source (TWALL) with three coal types for 9% O2 (Niksa and Liu, 2002). The transit time scale in this figure extends upstream of the heating element because radiation leakage and conduction along the flow tube preheat the suspension, and downstream of the reactor outlet, to cover the passage to a gas quench nozzle. The flow tube wall contains a fairly isothermal segment at 1400°C that delivers the heat flux that heats the particle suspension. The mean core gas temperatures seriously lag the particle temperatures through the first half of the transit time because the gas is transparent to radiation, and this stream is heated primarily by conduction off the flow tube wall through the sheath layer, and by conduction off the particle suspension. Consequently, the particle thermal histories are as much as 500°C hotter than the core gas during early times, but gas temperatures increase dramatically upon ignition and exceed the particle temperatures by the outlet (for this particular case). The thermal histories for different coals primarily differ because the heat release rates from char oxidation vary with coal quality, although variations in the thermophysical properties also matter. Due to the very hot tube wall and rapid rise in the gas temperatures, ignited chars approached 1600°C with the greatest inlet O2 levels. The equivalent reactor network in Fig. 6.3 was developed directly from CFD results according to the method in Chapter 5. Twenty-one CSTRs were used for the core flow in all cases, as a close approximation to plug flow. The average gas and tube wall temperatures as functions of transit time through the core were approximated as discrete SHEATH ZONE 3 - 9 CSTRs Sheath Gas

...

DEVOLATILIZATION ZONE 33 CSTRs Core Gas Char

...

COMBUSTION ZONE 17 CSTRs ...

Volatiles

Fig. 6.3 CSTR network for the core and sheath flows in RCFR tests at atmospheric pressure with different O2 levels. (Reproduced with permission from Niksa S, Liu G-S. Incorporating detailed reaction mechanisms into simulations of coal-nitrogen conversion in p. f. flames. Fuel 2002;81:2371–85. Elsevier.)

190

Process Chemistry of Coal Utilization

isothermal values for each CSTR in the series. The average gas and wall temperature histories were incorporated into an enthalpy balance that determined the thermal histories for the particles in suspension for both devolatilization and char oxidation. The only chemistry included for soot is its oxidation, whose kinetics were based on the closed-form rate expression for NSC kinetics. The predicted product distributions for primary devolatilization and secondary pyrolysis were then rendered into discrete injections of volatiles into each CSTR in the series that matched increments in the transit times to each CSTR residence time. In a similar way, sheath fluid was also injected into each of the core CSTRs to represent finite-rate entrainment. The entrainment rate was based on the nominal O2 utilization of 85% in the tests (Niksa and Cho, 1996), and the same mixing rate was imposed with all coals at all operating conditions. So the CSTR series was fed with char at the inlet, and injected with sheath fluid into every CSTR and with volatiles into only those CSTRs whose temperatures were hotter than the threshold for primary devolatilization. Additional details have been reported elsewhere (Niksa, 2019a; Niksa and Liu, 2002). Complete product distributions were recorded with uniform loadings of 0.15 gcoal/g-carrier with subbituminous, hv bituminous, and lv bituminous coals (Niksa, 2019a). For these three coals, the respective inlet O2 levels that would theoretically consume all volatile-derived products, including soot, were 7.9, 9.4, and 8.8%, respectively. The gas transit times were fixed at about 150 ms in all tests. Slip velocities for the 90 μm particles in these tests would move a soot cloud about 30 diameters away from its parent particle in the available transit time after ignition (Niksa and Cho, 1996). So envelope flames could not impose the environment for volatiles conversion in these tests. It is more likely that the gas phase was fairly well mixed. For each coal, the initial char oxidation reactivity was assigned to match the measured char yield with 14.6% O2 then used without adjustment to interpret the other test cases with less O2. None of these reactivity adjustments was out of the range of the expected sampleto-sample variability. The fixed fraction of char-N converted into NO was 0.1 for all coals and operating conditions. So no parameters were adjusted to improve the agreement with the measured product yields. As seen in Fig. 6.4, the ChemNet simulations accurately interpret the weight loss and the yields of soot, CO, CO2, H2O, and H2 across the domain of O2 level, and the predictions are generally within measurement uncertainties. The yields of C2H2 are underpredicted and those for CH4 are overpredicted for the richer conditions. The N-species distributions are qualitatively correct but subject to somewhat greater quantitative discrepancies. The char-N levels are overpredicted across the O2 domain, which explains the underpredicted HCN levels for rich conditions. However, the measured soot-N fractions in Fig. 6.4 imply N-percentages in soot approaching the coal-N percentage. But the bulk of reported measurements indicate soot-N percentages of only about half those in the parent coals (Niksa, 2018), so the predicted values may actually be closer to true values than these particular measurements. The product distributions for the subbituminous are similar, except for twice the extent of char burnout with 14.6% O2 and double the maximum amount of CO with 5.6% O2. The products with lv bituminous coal are almost completely unaffected by char oxidation because the maximum extent of burnout is only 10.5%. The maximum soot burnout is 16%, which is half the soot burnout with both other coals. Due to the

ChemNet furnace applications

191

Fig. 6.4 In clockwise order from the upper left panel, validation of weight loss and soot yields; yields of (dotted curve) C2H4, (□ and dashed curve) CH4, (● and dotted curve) C2H2, and (▪ and solid curve) H2; coal-N partitioning (excluding N2) as ( and dotted curve) char-N, (□ and dotdashed curve) soot-N, (● and dashed curve) HCN, and (▪ and solid curve) NO; and major gas intermediates and products with hv bituminous coal. (Reproduced with permission from Niksa S. Simulating volatiles conversion in dense burning coal suspensions. Part 1. Validation of reaction mechanisms. Fuel 2019a;252:821–31. Elsevier.)

very low char oxidation reactivity of the lv bituminous, the gas temperature history was significantly cooler than with both other coals, which diminished the soot oxidation rate. As the predicted product distributions are reasonably accurate across the domains of O2 and coal quality, the associated flame structures can be used to interpret the sequence of chemical stages responsible for NO production in primary flame zones. The flame structure in Fig. 6.5 for hv bituminous with 9.1% O2 conveys three distinct conversion time scales. The fastest burning fuels—C2H2, CH4, and H2—appear to be converted from 90 to 130 ms. However, somewhat less than half of these fuels were released from coal during this period. So GHCs and H2 continue to burn through most of the total transit time, but at lower concentrations than can be indicated in Fig. 6.5. The second major stage from 130 to 170 ms converts CO and HCN into CO2, N2, N2O, and NO, and consumes the bulk of char and O2. Roughly one-third of the N2 is due to char-N conversion, which was skewed by 90% toward N2 production in the ChemNet simulations (FNO ¼ 0.10). The remainder reflects HCN conversion in the gas phase, in part through an N2O intermediate. The third

192

Process Chemistry of Coal Utilization 0.40 16

hv bituminous 9.1% O2

14

0.35

Char

C2H2

0.30

0.20

8

Soot

0.15

6

0.10

4 2

hv bituminous 9.1% O2

CH4

0.05 0.00

0 9

O2

8 Concentration, %

Concentration, %

0.25

10

350

HCN

H2O

7 6

NO

300

N2

CO2

5

250

4 200

3 2

150

1.25 1.00 0.75 0.50 0.25 0.00

CO

hv bituminous 9.1% O2

100

H2 80

100

hv bituminous 9.1% O2

N2O 120

140

160 180 Time, ms

200

220

240

80

100

120

140

160 180 Time, ms

200

220

Concentration, ppm

Burnout, %

12

50 0

240

Fig. 6.5 Predicted flame structure with hv bituminous at 9.1% O2 as, in clockwise order from the upper left panel, extents of burnout of char and soot; the concentrations of CH4 and C2H2; and HCN, N2O, NO, and N2; and O2 and major gas intermediates (H2, CO) and products (CO2, H2O). (Reproduced with permission from Niksa S. Simulating volatiles conversion in dense burning coal suspensions. Part 1. Validation of reaction mechanisms. Fuel 2019a;252:821–31. Elsevier.)

and slowest conversion period extends from 150 to 230 ms when soot and additional char are converted at very low O2 concentrations. The CO released from the heterogeneous combustion promotes water gas shifting, which brings the H2 and CO concentrations back into the scale in Fig. 6.5. In turn, these species reduce a small portion of NO into N2 along the postflame region. With this particular coal, appreciable char oxidation coincides with the oxidation of all gaseous fuels and N-species conversion, whereas soot oxidation only promotes water gas shifting chemistry and minor extents of NO reduction. The influence of inlet O2 on the conversion of HCN into NO is presented in Fig. 6.6. HCN is converted over longer transit times with progressively less O2 and becomes incomplete for the two lowest O2 levels. The gas-phase in these cases is reducing due primarily to appreciable char burnout. Conversely, no NO forms at all with 2.9% O2, and NO is an intermediate species from 140 to 210 ms with 5.6% O2. NO production shifts toward markedly shorter transit times with both higher O2 levels while outlet NO concentrations are greater with more O2, as expected. As shown in Fig. 6.7, the impact of coal quality on NO production corresponds with two factors. First, the slower primary devolatilization rate with lv bituminous shifts HCN conversion into NO to longer transit times, whereas NO production starts at

ChemNet furnace applications

193

Fig. 6.6 Predicted HCN and NO concentrations with hv bituminous coal for (dashed curves) 2.9; (dot-dashed curves) 5.6; (solid curves) 9.1; and (dotted curves) 14.6% O2, where HCN and NO appear in red (dark gray in print version) and black, respectively. (Reproduced with permission from Niksa S. Simulating volatiles conversion in dense burning coal suspensions. Part 1. Validation of reaction mechanisms. Fuel 2019a;252:821–31. Elsevier.)

Fig. 6.7 Predicted concentrations of (left) CO and H2 and (right) HCN and NO where CO and HCN appear in red (dark gray in print version) and H2 and NO appear in black, for (dashed curves) subbituminous, (solid curves) hv bituminous, and (dotted curves) lv bituminous coals. (Reproduced with permission from Niksa S. Simulating volatiles conversion in dense burning coal suspensions. Part 1. Validation of reaction mechanisms. Fuel 2019a;252:821–31. Elsevier.)

similar times with both other coals because their devolatilization rates are only slightly different. The second and more important factor is the abundance and persistence of CO because HCN is present only as long as CO is present. This correspondence is especially clear in the concentrations with subbituminous which have two stages,

194

Process Chemistry of Coal Utilization

one for devolatilization and tar decomposition and another for char burnout, which was 40% with this coal at 9.1% O2. Both stages are clearly apparent in the concentration histories of CO, H2, HCN, and NO. Once HCN vanishes, NO saturates to its outlet level. The lv bituminous is predicted to give the greatest effluent NO concentration, by far, and the emissions from subbituminous are much lower than those from hv bituminous, as expected. The collection of chemical reaction mechanisms incorporated into these simulations accurately depict how coal quality and inlet O2 levels affect all the major reactants, intermediates, and products under the operating conditions in primary flame zones. The most reassuring aspect is that only a single parameter—the initial char oxidation reactivity—was adjusted to improve the quantitative agreement, but only for the greatest inlet O2 level with each coal. Otherwise, the dozens of other modeling parameters were fixed at the values assigned by the original mechanism developers or evaluated from default correlations with a coal’s proximate and ultimate analyses. The flame structures associated with these data interpretations reveal chemical stages that will reappear in primary flame zones at any scale: The gaseous fuels that burn in coal flames are the products of tar decomposition and secondary volatiles pyrolysis, not primary devolatilization products; they are mixtures of CH4, C2H2, H2, CO, HCN, and H2S, char, and soot. The fastest burning fuels in this reaction system are GHCs and H2 and the heat release from their combustion ignites primary flame zones. Tar decomposition into soot transforms a fast-burning hydrocarbon into the slowest burning fuel component of all. Considering that tar is often the most abundant primary devolatilization product, this transformation significantly diminishes heat release intensities. The bulk of the O2 is consumed in the simultaneous combustion of char, CO, and other noncondensable fuels. However, as char burning rates are substantially different for different coal types, the competition for O2 between char and gaseous fuels also largely determines how coal quality affects NO concentrations. This competition for O2 directly determines maximum NO concentrations, which arise when all HCN has been converted. Thereafter, the residual O2 is scavenged in the burnout of char and some soot which generates the CO that slowly reduces NO along the postflame to its concentration in the effluent. However, NO reduction in the postflame is only consequential when the transit time from the point of HCN elimination to the exit approaches 100 ms.

6.3.1.2 Coal flame structure at elevated pressures Does elevated pressure change the structures of primary coal flames? This is an especially difficult question to answer because diagnostic access becomes much harder, if not impossible, with very hot pressure vessels. The answer is nonetheless important because all entrained-flow gasifiers use primary flames in their first stage at substoichiometric SR-values. The study of coal flame structure at elevated pressures in this section was conducted in a pressurized version of the RCFR labeled as the “pRCFR.” One main difference with the RCFR for atmospheric pressure is that the particle flow is downward at atmospheric pressure and upward at elevated pressures, to counteract recirculation induced by buoyancy. Another is that the flow becomes

ChemNet furnace applications

195

transitional at 1.0 MPa and fully turbulent at higher pressures (vs. the downward laminar flow at 0.1 MPa). Turbulence is responsible for the most striking difference of all: Substantial fluctuations in radial velocities move coal particles from the core flow into the boundary layer on the tube wall, where they are unable to escape the strong shear field. Consequently, the coal suspension quickly moves off the centerline and into the sheath flow along the walls, which imparts a strong 2D character to the reaction system. Whereas the coal flame structures at atmospheric pressure were easily visualized in 1D, even the simplest flames of premixed coal suspensions with uniform inlet velocities at high pressure must be analyzed as a 2D system. Finally, the thermal capacitance of the gas flow is proportional to pressure so gas temperatures diminish for progressively higher pressures with uniform coal loading and the same radiant heat flux. Tests in the p-RCFR characterized one hv bituminous (Pit. # 8) at 1.0, 2.0, and 3.0 MPa and another (Ill. #6) at 1.0 and 2.0 MPa, and a subbituminous (PRB) at 1.0 MPa. Each test series contains from seven to ten individual tests with progressively higher inlet O2 levels, hence, SR-values. Throughout this case, SR-values are based on whole coal compositions unless noted otherwise. In the tests with Pit. #8 at 1.0 MPa, suspension loadings were nearly uniform at 4.7 wt%, whereas the inlet O2 mass fraction was varied from zero for the secondary pyrolysis case to 9.9% (SR ¼ 0.953). The suspension loadings were decreased from 4.7% at 1.0 MPa, to 2.3%–2.5% at 2.0 MPa, to 1.55% at 3.0 MPa; in other words, coal feed rates were essentially the same at all test pressures. Inlet O2 mass fractions were regulated at the higher pressures to impose similar ranges of SR-values in all test series. The maximum SR-values were near-unity with Pit. #8; 1.77 with Ill. #6; and 1.27 with PRB. As with the RCFR tests at atmospheric pressure, complete product distributions were monitored in individual runs of the p-RCFR (Liu and Niksa, 2005). Mass balances on individual tests closed within 2% for all cases while elemental balances closed within 5% or better. As in tests at atmospheric pressure, successive runs for progressively greater SR moved the system deeper through the chemical stages that consumed the different fuel components. Cases with SR less than one-half are dominated by the products of primary devolatilization, secondary volatiles pyrolysis, and partial oxidation of volatiles; whereas those with greater SR are governed by water gas shift equilibrium and the conversion of soot and char. When SR exceeded one-half, all hydrocarbon fuel components had been consumed within the furnace, but large amounts of CO and H2 persisted while soot levels diminished. The chars were fully ignited for all SR over 0.15, and extents of char burnout increased continuously for progressively higher SR, as expected. These tendencies are evident in the datasets for all three coals. The most significant differences among coals are the much higher soot yields for secondary volatiles pyrolysis of 21–23 daf wt% for both hv bituminous coals, and the tendency for fewer GHCs from coals of progressively higher rank. Char yields were similar at all pressures because even the lowest test pressure was higher than the threshold for asymptotic total weight loss for primary devolatilization. CFD simulations were developed for steady 2D axisymmetric flow (Liu and Niksa, 2005). Although several essential features of the flowfield became apparent in the CFD results, the original intent was to assign accurate thermal histories for subsequent

196

Process Chemistry of Coal Utilization

simulations with detailed chemistry. Accordingly, various parameters in the CFD chemistry submodel were freely adjusted to match the predicted and reported extents of consumption of gaseous fuels, soot, and char for each test. This procedure ensured that the simulated heat release profiles were accurate, which is important because the heat release profiles are the most uncertain aspect of heat transfer in this burner. In the CFD, core and sheath flows had equal flowrates but were distinguished at the inlet with a uniform particle suspension loading in the core and no particles in the sheath. Transport coefficients for heat and momentum transfer to the tube wall were assigned for transitional boundary layers at 1.0 MPa and fully developed turbulent flow at the higher test pressures. The κ-ε turbulence submodel and the two-layer zonal submodel for the near-wall treatment were implemented. This combination was the only one that accurately reproduced measured axial velocity profiles (Schlichting, 1979) as well as the turbulence intensity profiles across the boundary layers on vertical tubes from large-eddy simulations (LES) (Uijttewaal and Oliemans, 1996). In conjunction with the Stochastic particle dispersion submodel evaluated with 500 particle trajectories, the turbulent flow analysis was also qualitatively consistent with the particle concentration profiles in similar flowfields from LES (Uijttewaal and Oliemans, 1996). Turbulence intensities at the burner inlet were unimportant influences on the downstream flowfield. The radiant energy transfer within the burner was analyzed to assign the temperature profile along the flow tube and the radiant flux onto the flow axis. These profiles were incorporated into the CFD simulation as boundary conditions, along with convection from the flow tube through a developing boundary layer in either transitional or turbulent flow. Radiant fluxes onto the centerline of the 15.8 cm radiant section ranged from 61 W/cm2 at 3.0 MPa to 66 W/cm2 at 1.0 MPa and were within the range of 50–100 W/cm2 estimated for large pc flames. The measured extents of burnout for the various fuel components exhibit the expected sequence of fuel consumption from gaseous fuels to char to soot, with well-established tendencies for coal rank and pressure (Liu and Niksa, 2005). Considering the burner’s highly idealized geometry and inlet conditions, one could reasonably anticipate planar, nominally 1D flames across the central region of the flow tube, where the particle concentrations are highest. But such a simple structure is contradicted by the co-existence of substantial amounts of O2, CO, and H2 in the products from every test, except those for secondary volatiles pyrolysis without any O2. Typically, only 70% of the inlet O2 was consumed while about 30 daf wt% CO remained in the products, even for SR both well below and well above unity. Hindered penetration of O2 from the sheath flow into the core is an obvious explanation that turns out to be incorrect. The key is rapid particle dispersion off the centerline beyond the inlet. Almost immediately after injection, the particles acquire significant radial velocity components due to the turbulence and the wall collisions. All particles eventually penetrate into the sheath, and almost all of them contact the wall at some point. Once the particles move into the boundary layer, they are unlikely to escape back into the core flow so particles accumulate in the sheath and deplete the suspension loading in the core. Fig. 6.8 shows the radial profiles of normalized particle number concentration at six cross-sections from CFD. At the inlet, the particle concentration is uniform across

ChemNet furnace applications

197 18

8

14

6

12

5 %

10

4

8

3

6

2

4

1

2

0

0

8

0.001

0.002

0.003 R

0.004

0.005

0

0.006

0.002

0.003 R

0.004

0.005

0.006

0.003 R

0.004

0.005

0.006

0.003 R

0.004

0.005

0.006

16 14

5

12

4

%

10 8

3

6

2

4

1

2

0 0

0.001

0.002

0.003 R

0.004

0.005

0

0.006

0

0.001

0.002

30

9

6 cm

8

27 cm

25

7

20

6 5 %

4

15 10

3 2

5

1 0

0.001

18 cm

18

6

%

0

20

4 cm

7

%

12 cm

16

1 cm

7

0 0

0.001

0.002

0.003 R

0.004

0.005

0.006

0

0.001

0.002

Fig. 6.8 Radial profiles of particle number concentration of hv bituminous at 1.0 MPa and SR of 0.95 at axial positions of 1, 4, 6, 12, 18, and 27 cm. (Reproduced with permission from Liu G-S, Niksa S. Pulverized coal flame structures at elevated pressures. Part 1. Detailed operating conditions. Fuel 2005;84(12/13):1563–74. Elsevier.)

the core, which extends to almost 71% of the tube radius of 6 mm. At 1 cm downstream of the inlet, particles are still concentrated in the core, but a concentration gradient extends into the sheath. At 4 cm, the particle concentration profile has been inverted by dispersion into the sheath. Particle concentrations in the near-wall region continuously increase throughout the remainder of the burner while concentrations in the core diminish. Hardly any particles remain on the flow axis at the burner outlet. This result carries two important implications for the performance of the p-RCFR that were confirmed in the laboratory. First, particle agglomeration near and on the walls will likely cause operational problems due to the combination of slow particle

198

Process Chemistry of Coal Utilization

velocities, high particle temperatures, and high particle concentrations in the nearwall region. Second, complete O2 utilization will be difficult to achieve because of the strong accumulation of all fuels in the near-wall region vs. substantial amounts of residual O2 in the core. The segregation of the fuels and O2 in this flowfield imparts a strong 2D character to the flames. The most reactive gaseous volatiles ignite at the inlet to the radiant section. Their very high concentrations near the wall coincide with the hottest gas temperatures, which promotes convection that supplements the intense radiant flux from the tube. Consequently, particles dispersed into the sheath flow devolatilize before those in the core, and the gaseous fuel concentrations are correspondingly higher near the walls. Further downstream, the concentration spikes have been eliminated by combustion, and the highest gaseous fuel concentrations are within the core. The fuel inventory is then depleted by a flame front propagating from the sheath region toward the flow axis. This flame is fed by the diffusion of both fuel compounds and O2 from the core into the sheath flow. As long as the gas temperature remains cooler than the threshold for ignition, the core fuel concentration is reduced by transport into the sheath. But once the core ignites, the fuel is depleted too fast to escape the core. In light of the steep gradients in temperature, O2 concentration, and particle concentration in the CFD simulation (Liu and Niksa, 2005), it is not surprising that the particle RTD is broad and non-normal. Almost 80% of the particles have residence times between 330 and 480 ms, but the maximum time is 790 ms. The RTD has the form of a gamma distribution and resembles the RTDs for a few stirred tanks in series. As there are no short-circuits in this flowfield, the relatively few particles that become trapped in the wall boundary layer have significantly longer residence times than those remaining in the core flow. The mean residence time was about 450 ms with a standard deviation of 70 ms. The particle RTD is only one of the factors responsible for the broad distributions of the extents of char burnout from this burner. Another is that the particles in the core fully ignite and contribute to the highest levels in the burnout distribution only in tests with higher SR. Yet another is that burning particles in the sheath were extinguished by O2 depletion which freezes about half the population. These large variations in fuel consumption with SR affect gas temperature fields which, in turn, determine the 2D flame character, as seen in Fig. 6.9. The flame fronts were arbitrarily located on the locus of positions where gas temperature is 1050°C, which is certainly hot enough to ignite all fuel components. As the flow moves through the tube a flame front propagates toward the flow axis driven by convective heat transfer from the wall and by the heat released from the combustion of gaseous volatiles and soot. The flame is sustained by outward diffusion of volatiles and O2 toward the wall and by inward heat transfer toward the center. For some conditions, the annular flame front closes to a point on the flow axis; for others, the front cannot close before the end of the radiant section. The sketches of these flames share elements in common with both premixed Bunsen flames and laminar (Burke-Schumann) diffusion flames. But confined pressurized pc flames differ from both archetypes because fuel consumption is not restricted to the volatiles flame front. This flame segregates the flow according to the following three

ChemNet furnace applications

199

(a) S.R.=0.95

(b) S.R.=0.72

(c) S.R.=0.51

(d) S.R.=0.37

(e) S.R.=0.25

(f) S.R.=0.15

Fig. 6.9 Flame shapes for hv bituminous combustion at 1.0 MPa and at SR of (A) 0.95; (B) 0.72; (C); 0.51; (D) 0.37; (E); 0.25; and (F) 0.15. The position of the “flame front” is the locus of positions where the gas temperature is 1050°C. (Reproduced with permission from Liu G-S, Niksa S. Pulverized coal flame structures at elevated pressures. Part 1. Detailed operating conditions. Fuel 2005;84(12/13):1563–74. Elsevier.)

stages of combustion: (1) Gaseous volatile fuels and soot sustain the volatiles flame as it propagates from the near-wall region toward the flow axis; (2) Residual CO, H2, and char burn in the sheath flow after the volatiles flame has propagated deeper into the core; and (3) Within the core, residual gaseous fuels, soot, and char may eventually reach their ignition threshold and burn in a premixed mode. Whether or not the flame closes on the centerline in the available residence time will be mainly determined by pressure and SR. Inlet conditions that form closed flames at a lower test pressure will eventually sustain open flames at progressively higher pressures. None of the flames were closed for any coal at 2.0 and 3.0 MPa for SR values near unity. An equivalent reactor network was developed from the CFD with the guidelines in Chapter 5. The normalized combustibles flux defined in Eq. (5.1) delineates only two regions, as seen in Fig. 6.10. The greater particle concentrations near the wall are responsible for the greater combustibles mass fraction, compared to levels in the core. The combustibles mass fraction is uniform across a central core at all positions, except for a diminishing spike on the centerline that reflects the inlet condition. In the sheath flow, the combustibles mass fraction passes through a maximum with radial position,

Process Chemistry of Coal Utilization

Combustibles Mass Fraction

Combustibles Mass Fraction

Combustibles Mass Fraction

200

2.4 2.2 2 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 2.4 2.2 2 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 2.4 2.2 2 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0

Axial Location, cm

Pit. #8 PT=1.0 MPa

0 5 10 15 20 27.3

Axial Location, cm

Pit. #8 PT=2.0 MPa

0 5 10 15 20 27.3

Pit. #8 PT=3.0 MPa

0

0.1

0.2

Axial Location, cm 0 5 10 15 20 27.3

0.3 0.4 0.5 Radial Location, cm

0.6

0.7

Fig. 6.10 Profiles of combustibles mass fraction for hv bituminous at (upper) 1.0, (middle) 2.0, and (lower) 3.0 MPa. The bold solid and dashed lines on the right indicate how the radius that delineates the core and sheath regions is determined. (Reproduced with permission from Niksa S, Liu G-S. Pulverized coal flame structures at elevated pressures. Part 2. Interpreting NOX production with detailed reaction mechanisms. Fuel 2005;84(12/13):1575–85. Elsevier.)

ChemNet furnace applications

201

due to the accumulation of particles and the near-zero value on the wall. These portions of the profiles are very similar for all axial positions. As the turbulence intensity becomes stronger at higher pressures, the thickness of the region of higher combustibles mass fraction near the wall diminishes for progressively higher pressures. Profiles of combustibles mass fraction are similar for all axial positions, so only a single radial position delineates these regions. This radius was evaluated as the location of the intersection of lines through the combustibles fraction toward the boundary, as indicated in Fig. 6.10. The values are 0.45, 0.51, and 0.53 cm at 1.0, 2.0, and 3.0 MPa, respectively. The CSTR network for the baseline case of Pit. #8 at 1.0 MPa with an SR of 0.953 appears in Fig. 6.11. The networks for all other flames have similar branches and feedstreams but appreciably different quantitative specifications. Two CSTR-series appear in parallel, one each for the sheath (BL) and core (CR) flows. Both regions contain fuel particles, so both CSTR-series are fed with both entrainment gas and a portion of the entrained char suspension. In addition, entrainment gas from the core is gradually introduced into CSTRs for the sheath region at the rate of turbulent transport evaluated from the CFD simulation. Each region is subdivided into three axial zones. An attachment zone (not indicated in Fig. 6.11) simply conveys fuel components deeper into the burner because this entrance region is too cool to sustain any chemistry. The devolatilization zone covers the upstream portion of the reacting flow where volatiles are being released from fuel particles and burned with the entrained O2. The devolatilization times of 134 and 180 ms for the sheath and core, respectively, were determined from stand-alone devolatilization simulations for the average thermal histories of particles from the CFD simulation. The respective burnout zones appear downstream of the devolatilization zones where gas chemistry is minimal, but char continues to burn. The residence time in the char burnout zone of the sheath

BL DEVOLATILIZATION ZONE 10 CSTRs, 134 ms

BL BURNOUT ZONE 1 CSTR, 23 ms

...

...

...

...

Volatiles Gas Char

CR DEVOLATILIZATION ZONE 9 CSTRs, 180 ms

CR BURNOUT ZONE 1 CSTR, 5 ms

Fig. 6.11 Equivalent reactor network for a hv bituminous flame at 1.0 MPa and an SR of 0.953. (Reproduced with permission from Niksa S, Liu G-S. Pulverized coal flame structures at elevated pressures. Part 2. Interpreting NOX production with detailed reaction mechanisms. Fuel 2005;84(12/13):1575–85. Elsevier.)

202

Process Chemistry of Coal Utilization

region is 23 ms for this case, which sustains a significant extent of burnout; but the time available for char oxidation in the core is negligible in this and most other cases. The mean gas temperature in the sheath increased from 800°C to a maximum of 1550°C at 105 ms then cooled to about 1000°C at the tube exit. The gas temperature history for the core was much cooler over almost the entire reactor length but ultimately increased to 1530°C at the tube exit. Due to the significantly hotter temperature in the sheath, the residence time of 157 ms is shorter than the 185 ms in the core. In the CFD, the volatiles flame is closed. However, when flames do not close across the centerline the gas temperature history for the core remains much cooler even at the tube exit. Although the same radiation flux is imposed onto all particles, coal particles in the sheath flow were heated at 7600°C/s vs. 5500°C/s in the core, because the gas in the sheath is hotter. From 75% to 90% of the char was diverted into the near-wall region, where the upper end of this range pertains to cases at 1.0 MPa. Core fluid is entrained into the sheath CSTR-series as discrete entrainment increments for each CSTR, and the total entrained flowrate of O2 was deducted from the total O2 flowrate at the core inlet. Fig. 6.12 evaluates the predicted major products and the CO and H2 yields in the postflame gases with Ill. #6 at 1.0 MPa with the measured values at the p-RCFR exit. As SR was increased from 0.043 to 1.156, the measured weight loss due to devolatilization and char oxidation increased monotonically from 52 to 87 daf wt %. The CO2 yield increased from 1 to 190 daf wt% while the H2O yield increased from 2 to 44 daf wt%. The soot yield diminished from 22.5% to 3.7% as it was burned away in tests with the higher SR. As SR was increased, the CO yield passed through a maximum of 44 daf wt% at an SR of 0.5. The H2 yield decreased monotonically from 3.5 to 0.3 daf wt%. The ChemNet simulations depict the impact of SR on all major products except one within the measurement uncertainties. The weight loss and yields of CO2, H2O, H2, and soot are accurately predicted across the entire range of SR. The predicted CO yields are qualitatively correct but substantially lower than the measured values across the entire SR. range. This under-prediction is systematic for all coals in this test series. Notwithstanding, the predicted CO yields exhibit the correct functional dependence on SR. All the features in Fig. 6.12 are also apparent in the comparisons for all other test series (Niksa and Liu, 2005). In the results with Pit. #8 at 2.0 and 3.0 MPa, the weight loss and CO2 yields diminish slightly for progressively higher pressures, and the maximum in the CO yields becomes flatter and almost vanishes at 3.0 MPa. These same tendencies are seen in the results with Ill. #6 at 1.0 and 2.0 MPa, except that the CO yields versus SR are essentially the same at both pressures. Compared to the results for both hv bituminous coals at 1.0 MPa, the subbituminous had significantly more weight loss and higher CO2 yields at the highest SR, and the maximum CO yield was almost 20% higher. The most important impact of pressure is seen in the nitrogen speciation versus SR in Fig. 6.13. From left-to-right, the three upper panels show the impact of pressure increases with Pit. #8, and the first two lower panels show the pressure dependence with Ill. #6. Note the doubled scale for subbituminous at 1.0 MPa in the lower right corner. As expected from Fig. 6.4 for atmospheric pressure, with all coals at all

ChemNet furnace applications

203

250

III.#6, 1.0 MPa CO2

Major Products, daf wt, %

200

150

100

Weight Loss

50

H2O Soot

0 50

CO and H2, daf wt, %

III.#6, 1.0 MPa 40

30

20

10

CO H2

0 0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

S. R. Fig. 6.12 Evaluations of predicted (top) major products and (lower) CO and H2 yields for test series with Ill. #6 at 1.0 MPa. (Reproduced with permission from Niksa S, Liu G-S. Pulverized coal flame structures at elevated pressures. Part 2. Interpreting NOX production with detailed reaction mechanisms. Fuel 2005;84(12/13):1575–85. Elsevier.)

pressures the N-speciation shifts from a predominance of HCN at the lowest SR, toward minor amounts of NH3 at intermediate SR, toward rapid NO production at the highest SR. For all cases, N2 presumably becomes the major fixed-N species for all SR above 0.5. But two features for SR above 0.7 are striking and unanticipated: First, the conversions of coal-N into NO at all pressures are much lower than from the same fuels in similar flames at atmospheric pressure. For example, in tests in the

Fig. 6.13 Nitrogen speciation vs. SR for (top row) Pit. #8 at (left) 1.0, (center) 2.0, and (right) 3.0 MPa; and for (bottom row) Ill. #6 at (left) 1.0 and (center) 2.0 MPa. The lower right panel shows N-speciation for subbituminous at 1.0 MPa on an expanded scale. (Reproduced with permission from Niksa S, Liu G-S. Pulverized coal flame structures at elevated pressures. Part 2. Interpreting NOX production with detailed reaction mechanisms. Fuel 2005;84(12/13):1575–85. Elsevier.)

ChemNet furnace applications

205

RCFR at 0.1 MPa, subbituminous and Pit. #8 coals converted 21% and 16%, respectively, of their coal-N into NO at SR just over unity (Niksa and Liu, 2002). The comparable conversions at 1.0 MPa for both coals are well under 10%. More important, coal-N conversion clearly diminishes for progressively higher pressures with Pit. #8 and Ill. #6. With Pit. #8 it fell from 9.2% to 3.0% to 2.2% as pressure was increased from 1.0 to 2.0 to 3.0 MPa with SR near unity; with Ill. #6 the conversions fell from 8.7% to 5.3% for pressures of 1.0 and 2.0, respectively. The second striking feature in Fig. 6.13 is that HCN persists at SR values approaching unity at the highest test pressures. This persistence is stronger at 3.0 MPa than at 2.0 MPa with Pit. #8. In the ChemNet results for Pit. #8, the coal-N conversion to HCN is over-predicted at 1.0 and 3.0 MPa but is qualitatively correct throughout the SR range. The ultimate NO conversion is accurate for all pressures, falling off for progressively higher pressures. Ammonia is predicted and observed to be a minor intermediate, although the maximum in the NH3 levels is more pronounced in the predictions than in the measurements. Most importantly, the analysis predicts less conversion of coal-N to NO over the full range of SR for progressively higher pressures and the persistence of HCN at higher SR-values. For Ill. #6, the coal-N conversion to HCN at the lowest and the highest SR is predicted within experimental uncertainty for both 1.0 and 2.0 MPa. Ammonia levels are over-predicted throughout and exhibit maxima that are not apparent in the data. Nevertheless, the predicted NO levels are accurate across the full range of SR, and again indicate that coal-N conversion to NO diminishes for progressively higher pressures. For subbituminous, the coal-N conversion to HCN is slightly over-predicted for SR values below 0.85 but exhibits the correct qualitative form. The predicted NH3 levels are qualitatively correct but shifted toward higher SR by 0.30. However, the NO predictions are only qualitatively correct because the predicted NO level is double the measured value at the highest SR. The chemical structures of the flames in these tests provide context for the important finding that the conversion of coal-N to NO diminishes for progressively higher pressures. The predicted structure of the sheath region for the flame of Pit. #8 at 1.0 MPa appears in Fig. 6.14. In counterclockwise order from the upper left, the four panels display the histories of gas temperature and SR; the mass fractions of O2 and CO; the extents of burnout of char and soot; and the mass concentrations of the major N-species. In only this figure and Fig. 6.15, below, the SR-values do not include the combustibles in either soot or char and, therefore, indicate the oxidation potential for only the gas-phase chemistry. For this particular test, devolatilization is completed within 134 ms, and the flow leaves the reactor at 158 ms. The gas temperature approaches a maximum of 1540°C at 105 ms then decreases to 1150°C at the exit. The SR-value begins at infinity—because no fuel vapor is present at the inlet—then falls quickly while volatiles are released into the flow. But it never crosses the threshold for reducing conditions despite the abundant yield of volatiles from this coal, because more-than-half the volatiles are converted into soot, which does not factor into the SR-value in this figure. Even at the end of devolatilization, the SR is 1.24, which is still larger than the whole-coal-based value of 0.953. Clearly, the chemical environment in the sheath flow is much more oxidizing than expected

35

SR

1000 800

4

600 400

O2

O2, mass %

200 10

0

6

0.6

4

0.4

hv bituminous 1 MPa, SR = 1

2 0

0.2

25

50 75 100 125 Residence Time, ms

XSOOT

60

XCHAR

40

20

0

0.0 0

0

hv bituminous 1 MPa, SR = 1

80 0.8

CO

100

NH3

5

1.0

Entrainment Fraction

8

300

15

0

0 10

400

20

2

hv bituminous 1 MPa, SR = 1

200

SR HCN & NH3, ppmw

6

500

NO

25

Ent. Fr. & CO, mass % XCHAR & XSOOT, %

Gas Temperature, °C

8

1200

HCN

hv bituminous 1 MPa, SR = 1

30

TGAS

1400

NO, ppmw

10 1600

150

0

25

50 75 100 125 Residence Time, ms

150

Fig. 6.14 Chemical structure of the sheath of the flame with hv bituminous at 1.0 MPa and 0.953 SR showing, in counterclockwise order from the upper left, the operating conditions, major species, char and soot burnout, and N-species. (Reproduced with permission from Niksa S, Liu G-S. Pulverized coal flame structures at elevated pressures. Part 2. Interpreting NOX production with detailed reaction mechanisms. Fuel 2005;84(12/13):1575–85. Elsevier.)

1200 1000

30

800 20

600 400

TGAS

0.12

hv bituminous 1 MPa, SR = 0.95

80

40

SR

SR NO & HCN, ppmw

Gas Temperature, °C

100

50

hv bituminous 1 MPa, SR = 0.95

1400

0.10

NH3 0.08

60

NO 40

10

20

0

0

HCN

0

0.08 hv bituminous 1 MPa, SR = 0.95

0.02

0.04

2

CO

1

0

0

50

100 150 Residence Time, ms

0.02

0.00 200

CO, mass %

O2

Char & Soot Burnout, %

O2, mass %

3

0.00

hv bituminous 1 MPa, SR = 0.95

80

0.06

0.06 0.04

200

4

NH3, ppmw

1600

XSOOT

60

40

XCHAR

20

0

0

50

100 150 Residence Time, ms

200

Fig. 6.15 Chemical structure of the core region of the flame in Fig. 6.14. (Reproduced with permission from Niksa S, Liu G-S. Pulverized coal flame structures at elevated pressures. Part 2. Interpreting NOX production with detailed reaction mechanisms. Fuel 2005;84(12/13):1575–85. Elsevier.)

ChemNet furnace applications

207

from the whole-coal SR-value. Significant amounts of volatiles burn when the gas temperature is roughly 1300°C, based on the decay in the O2 concentration. All accumulated GHCs are consumed at ignition and neither H2 nor any of the hydrocarbon fuels are present in the sheath flow in significant amounts. GHCs ignite the flow but are otherwise unimportant. They are certainly not effective NOX reductants because NO forms well after they have been eliminated. The O2 concentration decays sharply during volatiles combustion and is depleted at about 120 ms. The CO concentration starts to increase sharply until all O2 is depleted, then reaches a maximum at 130 ms and decays during cooling. Its ultimate value reflects water gas shifting once all the O2 has been consumed. Char competes very effectively with the gaseous fuel compounds for the available O2 at the beginning, due to the very rapid burning rates of the smallest char particles in the PSD. The char ignites in the first reactor of the CSTR-series at 870°C. An ultimate char burnout of 67.1% is achieved at 120 ms when O2 is depleted. Despite its very small size, soot does not ignite until the gas temperature exceeds 1300°C, because of its low intrinsic oxidation reactivity. Soot oxidation soon competes with char oxidation for O2 because thermal annealing has retarded the char burning rate. Consequently, almost 90% of the soot in the sheath burns out. Again from Fig. 6.14, the NO concentration initially surges to 524 ppmw due to the rapid conversion of HCN in the boundary layer while the SR-value falls from 8 to 1.3. But once the available O2 falls below 2%, the NO concentration diminishes. The HCN concentration is much lower than NO’s and never exceeds 20 ppmw. It passes through two maxima at 40 and 134 ms. Ammonia appears as soon as NO reduction begins, but its concentration never exceeds 1.6 ppmw in the boundary layer. At the reactor outlet there is 187 ppmw NO but only 5 ppmw HCN and 0.2 ppmw NH3. The predicted structure of the flame core appears in Fig. 6.15. In this flame core, the devolatilization time is 180 ms, and the flow leaves the reactor at 185 ms. As in the sheath, neither H2 nor any GHCs are present in significant amounts. The gas temperature increases gradually from 50°C to a flame temperature of about 1500°C at the exit. The SR-value for the gas phase begins at infinity then falls sharply to 5 followed by a slow decrease toward 2.5, which is more-than-double the whole-coal value. The O2 concentration starts to decay at 100 ms when char particles ignite but does not vanish because there is insufficient fuel in the core. The flue gas O2 mass fraction is 1.5%. CO accumulates rapidly after 100 ms when char ignites. It approaches a maximum at 135 ms then decays through 160 ms. The CO concentration remains below 0.01% in the rest of the core. The ultimate extent of char burnout is 36 daf wt% vs. 86% for soot. HCN production begins at 40 ms and the HCN concentration grows to 85 ppmw at 120 ms. Whereas the O2 concentration during volatiles release is high enough to oxidize HCN into NO, the temperature is too cool. Ammonia appears as soon as HCN begins to decay, but its concentration remains well below 0.1 ppmw. HCN vanishes at 160 ms. NO does not form until 120 ms when the gas temperature is 670°C. As the core flow is always oxidizing, the NO concentration continues to increase to 66 ppmw at the exit, where NO and N2 (not shown) are the only fixed-N species. The ultimate NO level in the core is lower than that in the sheath by a factor of 3 because only 15% of the coal suspension flows through the core.

208

Process Chemistry of Coal Utilization

These flame structures generally pertain to other operating conditions and coal types, albeit with some significant quantitative differences. At the lower coal-based SR-values, the SR for the gas phase becomes as low as 0.5, which sustains more CO, H2, and GHCs in the effluent from the sheath. Similarly, HCN and NH3 become major N-species in the postflame gases. Extents of burnout for char and soot diminish over the entire flame. The flame cores release more CO but no GHCs, and small amounts of HCN and NH3 with the other fixed-N species in the postflame gases. The main difference for higher test pressures is that volatiles flames never close across the core flow at the highest test pressures, even for the highest SR. Consequently, the effluents contain more CO, HCN, NH3, char, and soot at progressively higher pressures. In the final stage of this analysis, an improved NOX production submodel for CFD was developed for pc firing at elevated pressures (Liu and Niksa, 2006). Based on the satisfactory evaluation in Fig. 6.13, a group of the detailed chemistry simulations was selected for sensitivity analysis to identify the major channels for fuel-N conversion, and to evaluate specific reaction rates and intermediate species concentrations. All such cases were for Pit. #8 in order to cover the widest possible ranges of pressure, temperature, and SR. The rate-of-production analysis indicated that HCN was converted into NO and N2 mostly through two parallel paths involving radicals of isocyanates (NCO) and amines (NHi), respectively. CN, HOCN, and HNCO only appeared as intermediates in these two major paths. N2 and NO were the major ultimate products of HCN-conversion whose relative yields depended on the gas environment. N2O and NO2 were minor products for the test conditions, so these species and their associated reactions were omitted from the global mechanism. The 3-step scheme in Fig. 6.16 is the same as a conventional NOX submodel except for the addition of two features: (1) The intermediate decomposition products of HCN—HNCO and amines—are explicitly represented as pseudo-HNCO; and (2) The additional concentration dependences on O2 and NO are also explicit in each of the three steps. The first addition is required to depict the high levels of residual HCN at moderate temperatures even for SR-values greater than unity. For progressively higher pressures at moderate temperatures, the oxyhydroxyl radical pool (O, OH, and H) shrinks, which decelerates the rate of HCN conversion into HNCO and amines. The second addition is needed to depict less NO production for

Fig. 6.16 Global scheme for HCN conversion into NO and N2 developed from sensitivity analysis of the ChemNet simulations for p-RCFR tests with hv bituminous. (Reproduced with permission from Liu G-S, Niksa S. A global NOX submodel for pulverized coal flames at elevated pressures. Combust Sci Technol 2006;178(5):953–74. Taylor and Francis.)

ChemNet furnace applications

209

progressively higher pressures during the early stages of coal conversion. While the oxyhydroxyl radical pool shrinks at elevated pressures, the conversion of HNCO shifts toward N2 production so the predicted NO levels are lower, consistent with the data. Indeed, the global scheme reproduces the ChemNet predictions for coal-N speciation as a function of SR in Fig. 6.13 across the entire pressure range, albeit with additional parameter adjustments (Liu and Niksa, 2006). The ChemNet analysis of pc flames at elevated pressures extends several of the main findings for coal flames at atmospheric pressure: The gaseous fuels that burn in coal flames are secondary pyrolysis products; GHCs and H2 burn fastest so they cannot possibly reduce NO; most O2 is consumed in the simultaneous combustion of char and CO; and the coal quality impacts are due to variable char oxidation rates. The most significant chemical difference at elevated pressure is the substantially lower conversion of coal-N into NO for progressively higher pressures for uniform SR. One might suppose that NO production is inhibited for the open 2D flame shapes in p-RCFR tests because of the strong thermal effects associated with the dispersion of coal particles in the sheath flow. But the basis in chemistry for inhibited NO production at elevated pressure was firmly established with CFD simulations that implemented the converged gas temperature field for 1.0 MPa in the cases for 2.0 and 3.0 MPa. In the associated ChemNet simulations with a uniform extent of char burnout, the fractional coal-N conversion to NO progressively decreased from 6.3% to 3.0% to 1.4% as pressure was increased from 1.0 to 2.0 and 3.0 MPa, which is comparable to the behavior in Fig. 6.13. Shifts in the gas phase chemistry are responsible for these tendencies whereby the decomposition of N-species intermediates (NCO, NHi) into N2 and NO favors N2 production at progressively higher pressures (Niksa and Liu, 2005). In broader terms, the analysis of pressurized pc flames also demonstrates ChemNet’s utility for 2D axisymmetric flowfields, even with highly nonuniform particle dispersion in transitional and turbulent flows. Provided that the transport coefficients are properly evaluated in the CFD simulations, the combustibles mass fraction is an effective means to transfer the 2D features into an equivalent reactor network. Neither very large temperature gradients nor exchanges of the flows through adjacent regions via entrainment pose severe problems. And the complexities in the flowfields only intensify in the case studies to come at larger scale.

6.3.2 PC combustion at pilot-scale The pair of cases for lab-scale in the previous section began a progression from coal flames in 1D to flames in 2D. But the more significant extension was from a flame with only a single region to one with two, by which effluent compositions were mixtures of disparate compositions from different regions with very different operating conditions. For larger flames, more regions come into play and more streams coalesce into the furnace flue gas. The cases in this section were developed for the single-burner pilot-scale furnace at the Combustion Research Facility (CRF) operated by Southern Research Institute. The furnace was fired at 1.75 MWth with one single-register burner. The burner is

210

Process Chemistry of Coal Utilization

weakly swirled, by design, to reproduce the trends in the NOX and LOI emissions from commercial T-fired furnaces operated by Southern Company Services, Inc. Indeed, these flue gas emissions mimic the variations with excess O2 and firing rate for commercial furnaces, and formal validations have been reported (Monroe et al., 1995). The goal of the test program was to characterize how biomass co-firing affects NOX and LOI emissions (Felix et al., 2003). Four coals representing most of the rank spectrum were cofired with two biomass forms, wood sawdust and switchgrass. Three biomass injection configurations were evaluated as co-milling into a single feedstream, co-injection into separate co-axial streams, and off-axis biomass injection into the primary coal jet, although only tests with co-milling are considered here. At least two biomass blend levels and three staging levels were tested for each fuel combination. The first subsection develops the equivalent reactor network for the CRF from commercial CFD simulations. The second uses ChemNet simulations of the coal-only flames to characterize their structures, the coal quality impacts on NOX emissions, and parametric sensitivities. A third describes how biomass co-firing alters the flame structures. A fourth subsection surveys NOX predictions for the entire database of fuel composition, biomass blend level, and staging level.

6.3.2.1 Reactor network for the CRF furnace The CRF furnace was designed to closely simulate the sequence of physicochemical processes in a full-scale utility boiler. The facility comprises coal crushing, milling, and classification; a 1.07 m ID, 8.5 m vertical, refractory-lined, water-cooled furnace; a single up-fired burner (single- and dual-register burners are available); a horizontal convective pass with three air-cooled tube banks; a series of heat exchangers, an ESP, a pulse-jet baghouse, and a packed-column SO2 scrubber. Separated OFA is injected through 4 off- radius ports located 4.6 m up the furnace. The staged test series was collected with 19% primary air, 66% secondary air, and 15% OFA. In the unstaged series, the OFA was combined with secondary air. Air flowrates were varied to produce flue gas (wet) O2 levels between 2.5% and 5%. The furnace handles gas velocities from 3 to 6 m/s, and has residence times from 1.3 to 2.5 s. The biomass was fed by co-milling coal with biomass into a single fuel feedstream through the burner core. The mean thermal history through the furnace and along the gas cleaning system compare closely to temperature data from commercial coal-fired furnaces (Monroe et al., 1995). Also, the CRF furnace was scaled for the same volumetric heat release ratio as commercial furnaces (evaluated as the ratio of the fuel heat input, in GJ/hr., to the radiant furnace volume, in GJ/hr-m3). The furnace was simulated, first, with conventional CFD to develop an equivalent reactor network and then with ChemNet simulations to generate flame structures and emissions predictions. The CFD simulations were performed by Reaction Engineering International (REI) with the chemistry submodel described in Chapter 4; the results included the specialized output described in Chapter 5. Due to the massive testing domain, relatively few tests were supported by CFD. Most CFD cases were for Pratt hv bituminous (PR) with and without biomass. Cases with subbituminous and lv bituminous were performed with REI’s Configurable Fireside Simulator (CFS) for the

ChemNet furnace applications

211

Fig. 6.17 Regions in the CRF tunnel furnace with weak swirl.

CRF. The CFS imposes a fixed computational grid on the calculations and is suitable for parametric case studies with the same firing configurations. No CFD results were available for cases with Galatia hv bituminous (GL) because these flames were expected to be very similar to the PR flames. The bulk flow patterns in the CFD results are sketched in Fig. 6.17. The particle trajectories show that all fuel particles remain close to the furnace axis through the NBFZ and throughout most of the upper furnace as well. Neither mixing in the NBFZ nor radial OFA injection disperses the particles off their original trajectories. Near the burner, the primary air stream is significantly expanded by the release of volatiles from the fuel suspension and by thermal expansion. This expansion zone delineates a fuel-rich core from the outer, swirled annular flow of secondary air. The expansion of the primary flow promotes short-circuiting of secondary air into the core because some of the secondary flow penetrates the expansion boundary. In addition, a portion of secondary air is entrained into the core as soon as it passes the edge of its delivery tube in the burner. Together, these entrainment mechanisms very rapidly mix about 20% of the secondary air into the primary flow. Nominal residence times in the core range from 120 to 170 ms. The flow of the core and secondary air streams gradually expands until it contacts the furnace wall midway to the OFA ports. A weak external recirculation zone (ERZ) forms in the corner bounded by the outer boundary of the secondary air stream. As the ERZ is too weak to entrain particles or appreciable amounts of air or fuel compounds, it was deemed to be inconsequential. As the fuel compounds in the flame core contact the secondary air stream, they mix and burn in an expanding mixing layer. This layer completely surrounds the core near the burner inlet and fills the entire furnace downstream of the core. The most distinctive feature of the mixing layer is that the temperature profile across the layer in a normal direction passes through a maximum value which is essentially the same around the entire circumference of the core. Note the similarity to the structure of a laminar diffusion flame, although flow in the CRF is definitely turbulent. Maximum gas temperatures approach 1700°C in cases where fuels of the highest heating values were fired without OFA and 1600°C in cases with 15% OFA. Residence times in the mixing layer to the OFA location vary from 500 to 600 ms.

212

Process Chemistry of Coal Utilization

The four air jets from the OFA ports do not penetrate onto the centerline. They also do not fill the entire flow cross-section. OFA injection does not impart significant swirl to the flow either, so sectors of the postflame gases mix independently with only one of the OFA jets. Downstream of the OFA ports the flow relaxes to a plug flow pattern that carries flyash and flue gas into the cleaning system. The total residence time to the furnace exit is approximately 2.5 s. More formally, distinct regions were delineated by ranges of the combustible mass fraction in Eq. (5.1) for the core, mixing layer, and ERZ, and by the O2 concentration for the SOFA zones. These flames comprise the following five regions of distinct chemical processes: (1) Core (CR)—Since the swirl is weak, the primary air and fuel stream remains intact as a flame core for a few meters, and this core retains virtually all fuel particles. (2) Mixing layer (ML)—Secondary air contacts the fluid from the core in a mixing layer that remains thin over most of the core length, but then fans out over the entire cross-section beyond the tip of the core. Almost all the secondary air mixes with the core flow downstream of the core tip. (3) ERZ—A relatively thin external recirculation zone fills the upstream furnace corners, pulling products from the mixing layer into the upstream secondary air stream. (4) OFA Zone—Downstream of the mixing layer, staged air is injected through four off-radius jets. Some eddies alter the gas flow streamlines near the injection ports, but particle trajectories are hardly affected by the OFA flows. (5) Burnout (BO) Zone—One and one-half seconds of residence time are available for the later stages of char oxidation downstream of the OFA injectors before the flue gas passes through a convective section and gas cleaning system.

In this flame, the core was delineated by the locus of points where the combustibles mass fraction equals 0.4, which is the minimum value that avoids convection of secondary air into the core at the inlet. The threshold for the boundary to the mixing layer is 0.3, which is the well-mixed value for the secondary air and primary streams. The OFA injection zone was delineated with an O2 mass fraction of 0.075. These regions were mapped in Fig. 5.4. These criteria were applied uniformly to all flames of all fuel combinations, configurations, staging levels, and SR. Indeed, the structures of all flames were similar in that they could be developed with the same regions. However, there are significant differences among the operating conditions specified for the same regions in different flames. Operating conditions through each region were evaluated with tracking of fluid and solid particles through the CFD fields, as described in Chapter 5. In fact, the RTDs for CRF flames from the fluid tracking were matched to CSTR-series for each region in Fig. 5.5. The RTD for the core was deconvoluted into one component for 16 CSTRsin-series and another for plug flow with respective mean residence times of 138 and 192 ms. The plug flow component represents the near-axial fluid motion under the influence of particle drag, and the CSTR-component represents flow with significant radial velocities. The RTD for the mixing layer was matched with a series of 19 CSTRs, and that for the OFA zone had 6 CSTRs-in-series. The burnout zone is essentially in plug flow. Mean thermal histories for core gas and coal (for devolatilization) were shown in Fig. 5.6, and the entrainment history for the mixing layer appeared in Fig. 5.7.

ChemNet furnace applications

213 NO REDUCTION ZONE 8 CSTRs, 73 ms

DEVOLATILIZATION ZONE 8 CSTRs, 65 ms Char Primary Air Volatiles Char Primary Air

...

...

DEVOLATILIZATION ZONE, 65 ms NO REDUCTION ZONE, 128 ms

MIXING LAYER 19 CSTRs, 508 ms

Secondary Air

...

...

Tertiary Air (OFA) OFA ZONE 6 CSTRs, 156 ms BURNOUT ZONE 1460 ms

LOI + Flyash Exhaust Gases

Fig. 6.18 Equivalent reactor network for the baseline PR hv bituminous flame.

The equivalent reactor network for these flames is represented by the one for the PR hv bituminous without biomass co-firing in Fig. 6.18. Networks for all other CRF flames have similar branches and feedstreams but appreciably different quantitative specifications. Only the four regions of the flame that contain fuel particles appear in Fig. 6.18. The ERZ was omitted because it is not strong enough in this furnace to entrain fuel particles and also because its extent was fairly small. The flame core has been subdivided into two regions. The devolatilization zone covers the upstream portion of the core in which volatiles are being released from the fuel suspension and burned with primary air. As the primary stream is reducing, very little residual O2 leaves the devolatilization zone. The NOX reduction zone covers the downstream portion of the core in which only the N-species are converted under the influence of water gas shifting due to the absence of O2. The CSTR-series for the mixing layer and the OFA zone represents the mixing of secondary and tertiary air streams, respectively. But there are no additional flows into the CSTR-series for the BO zone. Volatiles are entrained into the series for the first parts of the flame core; secondary air is entrained into the series for the mixing layer; and tertiary air is entrained into the series for the OFA zone. The addition rates of volatiles were specified from a standalone devolatilization simulation with the average thermal history of particles from the CFD simulation (in Fig. 5.6). The specific addition rates of the air streams were specified from the continuous entrainment profiles evaluated from the CFD simulation (in Fig. 5.7). Ranges for the regional residence times and CSTR-numbers are collected in Table 6.1. The core values have been subdivided further into zones for devolatilization and NOX reduction. Residence times for the entire core range from 100 to 170 ms. However, the more important specification is on the NOX reduction region because near-burner NOX emissions are largely determined by the time available for NO reduction before the primary flow penetrates the mixing layer (as explained below in Section 6.3.2.2). Cases with the shortest residence times in the NOX reduction zone

214

Process Chemistry of Coal Utilization

Table 6.1 Ranges of regional specifications. Region

Residence time, s

CSTR-number

Core - Devolatilization Core - NOX Reduction Mixing Layer OFA Jets Burnout Zone

0.060–0.080 0.040–0.110 0.530–0.575 0.055–0.110 1.400–1.800

5–12 7–20 12–25 6 15 or 35

tend to have significantly higher NOX emissions, all else being equal. The bulk flow pattern in the core is near-plug flow because the associated CSTR-series always contain more than 10 units and often had about 20 units. Residence times in the mixing layer were fairly similar at roughly 550 ms in all cases, and the flow pattern could be represented with CSTR-numbers similar to those in the core. The only region without a mostly plug flow pattern is the OFA zone. Six CSTRs were specified in all cases although the CFD-based RTDs often indicated CSTR-numbers as few as three. The nominal residence time for the burnout region is 1.5 s or longer. The flow pattern is strictly plug flow, so the simulations were initially conducted with 35 CSTRs. Succeeding cases gave the same results with 15 CSTRs because chemistry in the gas phase is negligible throughout this region. As the bulk flow patterns were very similar in all flames, variations in the network specifications on residence time and CSTR-number were not especially significant. But variations in the mixing intensities, especially in the mixing layer, were definitely important. Mixing constants varied from 6 to 9 s1 for the mixing layer in the coal-only baseline cases, and were roughly 9 s1 for the OFA jets. Values tended to be lower for co-firing cases. These variations significantly affect the predicted NOX emissions. Fourteen CFD simulations were available, in total, for the interpretation of over 300 test conditions. These simulations comprised one for three of the four coals with 15% OFA, plus one for one coal without OFA, plus two series on co-firing with both biomass forms with 15% OFA and various injection configurations. None of the CFD simulations characterized furnace stoichiometries other than that for 3.5% flue gas O2, and all the cofiring cases were for the high biomass levels. Network specifications were developed for all missing test cases by interpolations and, when necessary, extrapolations. Extrapolation was needed for different exit O2 concentrations and the low biomass cofiring level. To run cases with greater exit O2 levels, the flows of primary air, secondary air, and OFA were increased while the fuel feed rate was fixed. As the maximum gas and wall temperatures are always recorded in the mixing layer where secondary air mixes with the primary flow, gas and wall temperatures were increased by 5% for each 1% increase in O2 level with 15% OFA. With no OFA, the 5% increase was imposed when the O2 level was increased from 2.5% to 3.5%, but not when it was increased from 3.5% to 4.5%, to rectify systematic over predictions in trial simulations at the highest

ChemNet furnace applications

215

O2 levels. In addition, with 15% OFA, the fraction of secondary air diverted into the primary airstream at the burner outlet was increased from 0.18 to 0.20 to 0.22 as the exit O2 level was increased from 2.5% to 3.5% to 4.5%. With no OFA, only the first step change was imposed in the simulations. The absolute magnitudes are reasonably consistent with the CFD-values, which ranged from 0.14 to 0.21. Finally, the reciprocal time constant in the mixing model for the mixing layer was reduced by a factor of four in cases with 15% OFA at 2.5% O2, once it became apparent that the predicted NOX emissions based on temperature variations and diverted secondary air were consistently high. The adjustment factors for exit O2 and staging level are compiled in Table 6.2, where TB and RB denote the maximum ML temperature and mixing rate, respectively, for a baseline hv bituminous-only flame with 3.5% O2 and 15% OFA. None of the CFD simulations represented a low biomass cofiring level, which was either 5% or 10 wt% in the tests. Many of the simulation parameters for such cases were specified by interpolating values for the coal-only and high cofiring cases. The interpolated parameters included residence times in the core, mixing layer, and burnout zones; mixing parameters; and temperature histories. CSTR-numbers were usually the same for all fuel combinations. In general, biomass cofiring steepens the gas heating rate and slightly increases the maximum temperatures of the gases and walls. Consequently, both the magnitudes of the temperatures and the heating time scales were adjusted to obtain the interpolated temperature histories. The same temperature histories were applied to both biomass forms. The CFD-based regional residence times and mixing constants were used for each biomass form, although they were also very similar. The only adjustable parameters in the ChemNet analysis are the initial char oxidation reactivity and the fraction of char-N converted into NO during char burnout. Pairs of values for each coal were set to match the NOX and LOI emissions for coal-only flames with 3.5% exit O2 and 15% OFA. The assigned char reactivities were appreciably different than the CBK/E default values but within the expected ranges for each Table 6.2 Matrix of adjustment factors for exit O2 and staging. OFA, %

Exit O2, % 2.5

3.5

4.5

0.95TB (t) 0.18 RB/4 or 2.50

TB(t) 0.20 RB

1.05TB(t) 0.22 RB

TB(t) 0.18 1.5RB

1.05TB(t) 0.20 1.5RB

1.05TB(t) 0.20 1.5RB

15% OFA T(t) 20 Ent. Fr.a RMIX, s1 0% OFA T(t) 20 Ent. Fr. RMIX, s1 a

Entrainment fraction.

216

Process Chemistry of Coal Utilization

coal rank. Values for FNO varied from 0.22 to 0.48 which is also plausible. No further adjustments were made to improve the agreement among predicted and measured emissions levels. As biochars burn much faster than coal chars, a single default value for the biochar oxidation reactivity was used for both biomass forms, and FNO was zeroed out because char-N levels in biochars are usually very low. Each simulation with a network of 40 CSTRs took about 15 min on a microprocessor operating at 1.5 GHz and would take much less time with today’s much faster microprocessors.

6.3.2.2 Pilot-scale flame structure with coal The predicted structure of the flame core for the PR-only baseline flame appears in Fig. 6.19. In counterclockwise order from the upper left, the four panels display the variations in the gas temperature and SR values for the gas phase only; the mass fractions of O2 and CO; the extents of burnout for char and soot; and the mass concentrations of the major N-species. The SR values do not include the combustibles in either soot or char and, therefore, indicate the oxidation potential for the gas-phase chemistry. For this particular test, devolatilization is completed within 70 ms, and the flow leaves the core at 163 ms. Neither H2 nor GHCs are present in cores in significant amounts. The H2 mass fraction stays under 500 ppmw after the first 10 ms. Hydrocarbons are never present above

Fig. 6.19 Structure of the core of the baseline PR hv bituminous flame showing, in counterclockwise order from the upper left, the operating conditions, major species, char and soot burnout, and N-species.

ChemNet furnace applications

217

this threshold. GHCs ignite the flow but are otherwise unimportant. They are certainly not effective NOX reductants because NO forms well after they have been eliminated. The gas temperature increases rapidly during the first 25 ms then gradually approaches 1450°C at the core outlet. As the coal-based SR for the flame core (based on the flows of coal and primary air) is only 0.22, N-species could be expected to be converted under extremely rich conditions. In fact, the SR-value for the gas phase begins at infinity, which is the nominal value for pure primary air. It then falls sharply while volatiles are released into the flow, making it more reducing. It is as large as 1.7 while volatiles are burning. Then it diminishes during volatiles combustion to a steady level of 0.87, which is four times larger than the nominal value. This discrepancy is due to the incomplete conversion of coal into volatiles, the abundance of soot in the volatiles from this hv bituminous coal (which do not factor into the SR-value), and is counteracted by the consumption of the available O2 in the oxidation of soot and char. Clearly, the chemical environment in the core is much more oxidizing than expected from an SR based on the coal feed rate. The volatiles ignite at roughly 750°C based on the decay in the O2 concentration and the decay in the CO concentration. At this point, two-thirds of the ultimate volatiles yield has been released. All accumulated GHCs are consumed at ignition, and GHC concentrations remain very low throughout. The O2 concentration decays sharply during volatiles combustion, then decays more gradually after the char and soot ignite at 25 ms. The CO concentration decays during the ignition period then gradually increases during the oxidation of char and soot. Its ultimate value reflects water gas shifting once all O2 has been consumed. Char competes very effectively with the gaseous fuels for the available O2 in the core, due to the very rapid burning rates of the smallest char particles in the PSD. The char ignites when the gas temperature is 929°C and loses almost a third of its mass before the annealing mechanism in CBK/E comes into play. Subsequent burnout is much slower due to the combined influences of annealing and O2 depletion. Even so, almost half the char burns out in the core. Despite its very small size, soot is much harder to bring to the fully ignited state because of its low intrinsic oxidation reactivity. Consequently, only 11% of the soot burns out in the core. The N-conversion chemistry expected for the coal-based SR-value does not materialize within the flame core. The NO concentration initially surges to 670 ppmw due to the rapid conversion of HCN, the primary volatile-N species, in the lean section of the core, where the SR-value falls from 5 to unity. But once the available O2 falls below 5%, the NO concentration diminishes in tandem with the decaying HCN concentration. Ammonia appears as soon as NO reduction begins, but its concentration never exceeds 16 ppmw in this core. The NO reduction stage (during which no additional volatiles are released) coincides with the second surge of NH3 and with the final rapid decay in the HCN concentration. At the end of the core, there is 244 ppmw NO but only 5 ppmw HCN and 1 ppmw NH3. Profiles through the mixing layer from the PR-only baseline flame appear in Fig. 6.20. This region has the hottest temperatures in the flame, due to the rapid mixing of secondary air with combustibles from the core. The gas temperature history exhibits the peaked profile expected for flames of segregated fuel and air streams. It reaches its maximum value of 1550°C in just under 300 ms, then gradually diminishes over the remaining

218

Process Chemistry of Coal Utilization 1.5

275

1.4

HCN

1500 1.2

SR NO, ppmw

1.3

4

225 2

200

1475

NH3

1.1

hv bituminous Mixing Layer 15 % OFA, 3.5% O2

1450 2.0

hv bituminous Mixing Layer 15 % OFA, 3.5% O2

175

1.00

Ent. Fr.

0 100

O2

XSOOT

0.75

CO

1.0

0.50

0.5

0.25

Ent. Fr. & CO, mass %

O2, mass %

NO

250

200

300

400

500

Residence Time, s

600

0.00 700

80

60

XCHAR

40

hv bituminous Mixing Layer 15 % OFA, 3.5% O2 0.0

NH3 & HCN, ppmw

SR

1525

1.5

6

hv bituminous Mixing Layer 15 % OFA, 3.5% O2 200

300

400

500

600

XCHAR & XSOOT, %

1550 Gas Temperature, °C

300

TGAS

20

0 700

Residence Time, s

Fig. 6.20 Structure of the mixing layer of the baseline PR hv bituminous flame showing, in counterclockwise order from the upper left, the operating conditions, major species plus entrainment fraction of secondary air, char and soot burnout, and N-species.

400 ms. The SR history closely follows the cumulative entrainment fraction, increasing with the addition of secondary air from close to unity to just under 1.4. Eventually, the SR-values diminish due to the burnout of char and soot, which tends to pull the SR-values toward unity. But the accumulation of O2 in the layer is impeded by its rapid consumption during soot oxidation. While soot is present, the O2 concentration rises to 0.4% by mass. Then after most of the soot has burned out, the O2 concentration rises to 1.8% during the last quarter of char burnout. As the gas phase becomes more oxidizing, the CO concentration diminishes and ultimately vanishes after the soot has burned away. The gas temperature into the mixing layer exceeds the threshold for rapid ignition of soot, and the soot burning rate accelerates while the gas temperature increases. The soot is completely burned out when the gases reach their maximum temperature. Char burns much slower than soot in the mixing layer, in contrast to the order of their burning rates in the cooler flame core. Both the higher temperatures in the mixing layer and the stronger annealing mechanism in the char oxidation rate are responsible for the reversal. The N-species chemistry is only interesting at the inlet to the mixing layer. Both HCN and NH3 are eliminated very quickly, continuing the tendency for NO production established late in the flame core. The NO concentration initially falls due to the addition of secondary air. It then rises across the trailing two-thirds of the mixing layer

ChemNet furnace applications

219

while the conversion of char-N into NO supplements the NO inventory. Since there are no reducing agents in the rest of this region, gas-phase chemistry has been decoupled from char-N conversion. In fact, chemistry in the gas phase remains unimportant throughout the remainder of this furnace. The only chemistry in the OFA and burnout zones is char oxidation, with simultaneous NO production from the conversion of char-N. As seen in Fig. 6.21, the O2 concentration surges to almost 4.5% during the addition of OFA then diminishes to 4% during the latest stages of burnout. The ultimate O2 concentration equals 3.9% by volume, close to the 3.7% in the corresponding test. The gas temperature is quenched by 150°C during OFA injection, then gradually cools to 1150°C at the furnace exit. The extent of char burnout asymptotically approaches 100% across the OFA and burnout zones. The burning rate diminishes throughout due to the

1500

1.0

Ent. Fr.

hv bituminous OFA & BO 15 % OFA, 3.5% O2

0.6 1300 0.4

TGAS

0.2

1200

4.5

Entrainment Fraction

Gas Temperature, °C

0.8 1400

O2

0.0 100

XCHAR

95

3.5

XCHAR, %

O2, mass %

4.0

3.0 90

2.5 2.0

hv bituminous OFA & BO 15 % OFA, 3.5% O2

1.5

85 1.0

1.5 Residence Time, s

2.0

Fig. 6.21 Structure of the OFA and BO regions of the baseline hv bituminous flame showing (top) the operating conditions and (bottom) O2 concentration and char burnout.

220

Process Chemistry of Coal Utilization

100

Extent of Burnout, %

80

BO

60

40 OFA 20 ML

hv bituminous 15 % OFA, 3.5 % O2

CR

0 0

50

100

150 200 Particle Size, µm

250

300

350

Fig. 6.22 The extent of char burnout versus char size at the exits from the core, mixing layer, OFA region, and burnout region.

combined influences of the falling gas and wall temperatures, annealing during the previous thermal treatments, and, perhaps, ash encapsulation. The ultimate extent of burnout is 99.1%, which corresponds to an LOI value of 2.0 wt%. This prediction is almost double the measured value of 1.2%. The predicted NO concentration across the OFA region falls from 287 to 270 ppmw, then rises to 290 ppmw across the BO region. The corrected flue gas NO concentration of 335 ppm @ 3% dry O2 essentially equals the measured value of 328 ppm for this test. Extents of char burnout are resolved over the char PSD at the exits of the four regions of this flame in Fig. 6.22. Only particles smaller than 150 μm have ignited by the end of the core, yet the consumption of O2 by char oxidation is significant because particles smaller than 60 μm have mostly burned out. The entire PSD has ignited by the end of the mixing layer. But the acute size dependence persists through the end of the OFA region. By the end of the burnout region, particles smaller than 150 μm have completely burned out so only the largest char particles contribute to UBC and LOI. For CRF-flames, which do not disperse or recirculate particles off the furnace centerline, this analysis predicts that UBC comprises the remnants of only the largest particles in the char PSD.

6.3.2.3 Pilot-scale flame structures with biomass cofiring The predicted structure of the flame core for the PR/20% switchgrass (SG) flame appears in Fig. 6.23. Even though the SG has much more fuel-N than PR coal (2.4 vs. 1.8 daf wt%), the cofired flame generates 40% less NOX in the flue gas. So the structure of this flame illustrates how biomass affects the flame chemistry to reduce

ChemNet furnace applications

221 25

1200 1000

HCN

15

800

600

NO

10

600 400

hv bit/20% SG Core 15 % OFA, 3.5% O2

200

400

5

200

NH3 0 2.0

O2

hv bit/20% SG Core 15 % OFA, 3.5% O2

CO

20

0 50

XCHAR

15 1.0 10

hv bit/20% SG Core 15 % OFA, 3.5% O2

5

H2

0 0

25

50

75 100 125 Residence Time, s

0.5

CO & H2, mass %

1.5

30

20

10

XSOOT 0.0

150

175

40

XCHAR & XSOOT, %

0

O2, mass %

1000

800

SR

Gas Temperature, °C

20

NO, HCN, & NH3, ppmw

hv bit/20% SG Core 15 % OFA, 3.5% O2

1400

0 0

25

50

75 100 125 Residence Time, s

150

175

Fig. 6.23 Structure of the core of the PR/20%SG flame showing, in counterclockwise order from the upper left, the operating conditions, major species, char and soot burnout, and N-species.

NOX. The gas temperature increases at a slightly slower heating rate than in the PRonly flame, primarily because SG contains eight times more moisture than PR coal. As expected, the SR-value for the gas phase falls sharply while volatiles are released into the flow, making it more reducing; it then relaxes to an ultimate value of 0.951. The gas phase is more reducing than the PR-only flame core because there is much less soot and more volatiles from SG. The volatiles ignite at roughly 650°C, based on the decay in the O2 concentration. The O2 concentration decays sharply during volatiles combustion then decays more gradually after the char and soot ignite at 25 ms. As for the baseline PR flame, GHCs ignite the flow but are otherwise unimportant. But the H2 mass fraction persists at roughly 1000 ppmw across the entire core, and there is twice as much CO. Initially, the CO concentration surges during the ignition period then increases more gradually during the oxidation of char and soot. Its ultimate value and the persistence of H2 reflect water gas shifting once all O2 has been consumed. Almost half the char but only 5% of the soot burn out in the core, which are similar to the burnout in the PR-only flame. Whereas the macroscopic combustion characteristics discussed to this point are very similar in the cofired and baseline flames, there are crucial differences in the N-species conversion chemistry. Most importantly, all three of the major fixed-N species are present at similar concentrations. Ammonia is expelled by SG to a maximum concentration of 729 ppmw. But it is converted to HCN in only 40 ms, demonstrating

222

Process Chemistry of Coal Utilization

that the primary forms of volatile-N are unimportant because gas-phase chemistry governs the N-speciation in a flame core. The distribution of N-species is determined by the SR-value in the gas phase which, in this flame, is too lean to sustain much NH3. The HCN concentration surges above 830 ppmw then decays in tandem with NH3 while the NO concentration grows to 620 ppmw. Thereafter, all the fixed-N species concentrations remain fairly constant before they plummet at the exit of the core. Although the exit NO level from this core is the same as from the PR-only flame core, the HCN level is much higher at 350 ppmw. When the gases from the core contact secondary air in the mixing layer, the NO concentration relaxes to 170 ppmw by the time that both the other fixed-N species have been eliminated, which is essentially identical to the 179 ppmw NO at this same condition in the PR-only flame. Beyond this point, gas-phase chemistry becomes inconsequential. But the remaining inventory of char-N in the biomass cofired flame is much lower because 68% of the PR-char had burned out when NO became the only fixed-N species, versus 53% in the PR-only flame. The fact that the residual SG-char contains no nitrogen compounds this difference. The reason that the exit NOX emissions are lower for the SG-cofired flame is that the fuel blend releases a much larger portion of the fuel-N into the core, which then becomes subject to NO reduction in the gas phase chemistry. This enhancement was due to two independent factors: First, PR-char competes more effectively for the available O2 than SG-char, due to its smaller size, so the extent of burnout for PR-char is higher in the cofired flame than in the PR-only flame. As the extent of burnout increases, the amount of residual char-N diminishes. Second, SG completely releases its fuel-N during devolatilization and, therefore, does not convey any char-N into the downstream regions of the flame. Higher concentrations of HCN and NH3 in the core promote NO reduction. Moreover, the higher levels of CO and H2 associated with the lower SR-value promote NO reduction in the core as well as in the inlet to the mixing layer. Even though more NO can be produced early in the core when more volatile-N species are released, the cores in CRF-flames provide sufficient residence times to reduce it away. However, the shorter residence time in the core of the SG-cofired flame did not enable all the HCN to be eliminated, as it was in the PR-only flame. Perhaps the NO emissions could have been reduced further if there was some way to extend the transit time through this core. The competitive burnout of the PR- and SG-chars is apparent in Fig. 6.24. The predicted extents of burnout are plotted versus the sizes in the initial PSDs of these chars. PR-particles smaller than 160 μm have ignited by the end of the core. Slightly larger SG-particles have also ignited because the initial char oxidation reactivity of biomass is much faster than any coal’s reactivity. But by the end of the mixing layer, PR-char smaller than 120 μm burns faster than this portion of the SG-PSD. The reactivity reversal reflects the huge ash content in SG-chars (76%), which is 2.5 times greater than PR’s. According to CBK/E, the ash forms a layer around the combustibles that inhibits O2 penetration and thereby reduces the burning rate. By the end of the OFA region, the entire PR-PSD burns faster than all the SG-PSD. And at the furnace

ChemNet furnace applications

223

100

BO

hv bit/20% SG 15 % OFA, 3.5% O2

Extent of Burnout, %

80 CR 60

OFA

40 ML 20

0 0

50

100

150

200

250

300

Particle Size, µm

Fig. 6.24 The extent of char burnout of (solid curves) hv bituminous char and (dashed curves) switchgrass char vs. initial char size at the exits of the CR, ML, OFA, and BO regions in the PR/20%SG flame.

exit, essentially all the predicted UBC is remnants of the largest SG-particles. The predicted LOI emission of 1.9 wt% compares well with the reported value of 1.6%. The predicted structure of the flame core for the PR/20%SD flame appears in Fig. 6.25. In contrast to SG, SD has almost no fuel-N. However, this cofired flame still generates 40% less flue gas NOX than the PR-only baseline flame, which is double the reduction expected from the removal of fuel-N alone. So the structure of this flame reflects distinctive chemistry that reduces NO within the NBFZ. The time scale in Fig. 6.25 has been extended to depict the core as well as the first third of the mixing layer. For this particular test, devolatilization is completed within 80 ms, and the flow leaves the core at 118 ms. The total residence time in the mixing layer is 476 ms, but only the first 150 ms are shown in Fig. 6.25. The SR-value for the gas phase falls sharply while volatiles are released into the flow, making it more reducing. It then relaxes to an ultimate value of 0.864, which is significantly more reducing than the PR/20%SG flame. Although SD and SG have identical volatiles yields, the gas phase in the SD-core becomes more reducing because more CO and less soot are produced by the primary volatiles from SD. Initially, the CO concentration surges during the ignition period, then increases more gradually during the oxidation of char and soot. Its ultimate value and the persistence of H2 reflect water gas shifting once all O2 has been consumed. The maximum CO concentration is double that in the PR/20%SG flame. The H2 mass fraction persists at roughly 1000–2000 ppmw across the entire core. Moreover, GHCs, especially C2H2 (not shown), persist at 1000 ppmw or more across the entire devolatilization zone. This is the first flame core with an appreciable level of GHCs in the presence

224

Process Chemistry of Coal Utilization 1600

15

800 10

600

hv bit/20%SD CR & ML 15 % OFA, 3.5% O2

0

CO

NO

800

80 60

600

NH3

400

40

200

20

5

0

hv bit/20%SD CR & ML 15 % OFA, 3.5% O2

0

hv bit/20%SD CR & ML 15 % OFA, 3.5% O2

80

XSOOT 0.8

3

15 2 10 1

5

CO & H2, mass %

O2, mass %

20

120 100

1000

0

O2

140

60

XCHAR

Ent. Fr.

0.6

40

0.4

20

0.2

Entrainment Fraction

SR

200

SR NO & HCN, ppmw

1000

hv bit/20%SD CR & ML 15 % OFA, 3.5% O2

1200

XCHAR & XSOOT, %

Gas Temperature, °C

20

1200

400

HCN

1400

NH3, ppmw

25

TGAS

1400

H2 0 0

50

100

150

200

Residence Time, s

250

0 300

0 0

50

100

150

200

250

0.0 300

Residence Time, s

Fig. 6.25 Structure of the core and mixing layer of the PR/20%SD flame showing, in counterclockwise order from the upper left, the operating conditions, major species, char and soot burnout, and N-species.

of NO, although their concentration is still much lower than those of CO and H2. As for the other flames, almost half the char burns out in the core but hardly any soot burns out in this particular core. The more reducing character of the gas phase in this core imparts several distinctive features to the N-species conversion chemistry. The N-speciation is dominated by HCN and NO, as for the PR-only baseline flame. The NH3 released by the SD is rapidly converted into HCN and NO within 40 ms. But NO does not accumulate at the expense of HCN, as in both of the other flame cores. Instead, the HCN concentration surges while NO accumulates. Some prompt N-fixation mechanisms involving N2 in the air must be responsible because the total maximum amount of fixed-N species is double the maximum value in the PR/20%SG flame, and the SG-cofired flame has significantly more volatile-N. NO reduction begins at 75 ms but by the exit of the core, there is still 1390 ppmw HCN and 465 ppmw NO, which are both much higher than in either of the other flames. Very early in the mixing layer, NO reduction accelerates while the HCN concentration plummets. The NH3 concentration reaches 92 ppm before vanishing with the HCN concentration. The ultimate NO concentration after both other fixed-N species have been eliminated is only 143 ppmw, which is 25 ppmw lower than the analogous level in the PR/20%SG flame. Even though there was a higher concentration of

ChemNet furnace applications

225

100

BO

Extent of Burnout, %

80

hv bit/20% SD 15 % OFA, 3.5% O2

CR 60

40

OFA

20

ML

0 0

50

100

150

200

250

300

Particle Size, µm

Fig. 6.26 The extent of char burnout of (solid curves) PR hv bituminous char and (dashed curves) SD char vs. initial char size at the exits of the CR, ML, OFA, and BO regions in the PR/20%SD flame.

fixed-N species in the core, the greater reducing potential yielded a lower NO concentration in the mixing layer after chemistry in the gas phase was exhausted. As the extents of char burnout at this point are comparable for the PR/20%SG and PR/ 20%SD flames, the 20 ppmw reduction for the PR/20%SD flame persists in the flue gas emissions as well. The extent of soot burnout surges with the entrainment of secondary air. Due to the high maximum temperatures in the mixing layer, the extent of soot oxidation eventually overtakes the extent of char oxidation at 220 ms. The soot burns out in the mixing layer, whereas char is carried over into the OFA and BO regions. The extents of burnout of the PR- and SD-chars in Fig. 6.26 show that this biomass char wins the competition for O2, in contrast to the ash-laden SG-char. There is only 2.5% ash in the SD-char, versus 30.9% for the PR-char. So the advantage of the biomass’s faster initial reactivity persists throughout the entire furnace. Indeed, the UBC from this flame is predicted to consist entirely of fragments of unreacted PRchar, which is the opposite of the character of the LOI emissions from the PR/20% SG flame.

6.3.2.4 NOX predictions for pilot-scale flames with biomass cofiring This subsection evaluates predicted NOX emissions over the test domain, including all combinations of the four coals and two biomass forms at two biomass cofiring levels. The fuel quality impacts are largely determined by fuel compositions and, especially,

226

Process Chemistry of Coal Utilization

Table 6.3 Fuel properties. Switch Grass SG

Jacobs Ranch JR

Galatia GL

Pratt PR

Jim Walters JW

Moisture 9.5 Ash 0.4 Volatile 78.1 Matter Fixed 11.9 Carbon Ultimate, daf wt%

15.2 29.5 47.6

19.3 5.4 39.6

5.8 6.6 33.7

1.9 15.1 33.2

0.8 14.6 20.0

7.7

35.7

54.0

49.9

64.6

C H O N S PSD

49.8 6.1 43.9 0.2 0.0

56.1 5.4 35.7 2.4 0.4

74.9 4.9 18.8 0.9 0.4

81.8 5.0 10.0 2.0 1.1

83.4 5.5 7.5 1.8 1.8

89.5 4.6 3.3 1.7 0.9

, μm RR-n RR-b, cm

163 2.1808 5518.2

173 2.2617 6728.2

28.8 0.9405 169.71

53 1.7353 6116.5

48 1.3111 752.06

34 1.0082 211.46

Sawdust SD Proximate, as rec’d

by the product distributions for primary devolatilization and tar decomposition. These aspects are reviewed before the formal evaluation of the NOX predictions. The fuel properties are collected in Table 6.3 in order of increasing rank from left to right. Moisture levels are highest for the low-rank fuels, especially SG and JR. Ash levels are widely variable and especially high for biomass SG and for coals PR and JW. Whereas it appears that the volatility of SD is much higher than SG’s, on a daf basis, their volatilities are almost identical. However, the daf volatiles contents of the coals fall by more than a factor of two over this suite of samples, which will definitely affect the conversion of coal-N into NOX. Carbon contents increase and oxygen contents decrease for fuels of progressively higher rank. The pair of biomass samples represents most of the range of elemental compositions seen for diverse forms of biomass. The represented range of coal rank, from subbituminous through lv bituminous, is similarly broad. The hydrogen, nitrogen, and sulfur levels are not rank-dependent. Whereas almost all biomass contains little nitrogen and sulfur, sample SG contained the most nitrogen of any of the fuels, due to its decomposition before firing. The PSDs are typical utility grinds for the coals, except for the much finer grinds of JR and JW, whereas the biomass grinds are much coarser, as expected. The pairs of Rosin-Rammler parameters were assigned from pairs of measured size fractions. The predicted distributions of secondary pyrolysis products are collected in Table 6.4. Total volatiles yields are the same for both forms of biomass and, at 86 daf wt%, are much greater than the yields from any of the coals. The biomass product

ChemNet furnace applications

227

Table 6.4 Compositions of secondary pyrolysis products and chars. SD

SG

JR

GL

PR

JW

Wt. Loss Soot CH4 C2H2 C2H4 H2 CO CO2 H2O HCN NH3 H2S Char Comp., daf wt%

86.1 4.3 7.1 2.2 1.4 2.1 48.4 8.2 12.1 0.0 0.24 0.0

86.0 13.4 7.4 1.3 1.5 1.7 41.5 8.0 7.7 0.0 2.90 0.44

65.2 30.1 0.7 1.5 0.0 3.4 12.9 6.4 7.6 1.26 0.0 0.42

56.5 33.7 0.4 1.0 0.0 3.6 7.2 2.2 4.9 2.47 0.0 1.17

59.8 37.9 0.5 1.3 0.0 4.0 6.1 1.7 4.3 2.24 0.0 1.91

39.7 26.8 0.3 2.3 0.0 3.5 1.7 1.0 1.8 1.52 0.0 0.96

C H O N S Char ash, wt% Char size, μm

94.7 3.4 1.9 0.0 0.0 2.5 97.6

97.1 2.5 0.4 0.0 0.0 75.7 103.6

98.9 0.5 0.0 0.4 0.1 15.9 29.9

98.4 0.5 0.0 1.1 0.1 14.2 59.1

98.5 0.4 0.0 1.0 0.1 30.9 54.1

98.2 0.5 0.0 1.26 0.0 21.8 41.7

Volatiles, daf wt%

distributions are dominated by CO, with substantial amounts of GHCs, especially CH4, and CO2 and H2O. Hydrogen is another significant fuel compound from biomass. But there is relatively little soot, considering that tar (a.k.a bio-oil) is the soot precursor and accounts for 25% to 45% of the daf fuel mass released during primary devolatilization. The reason is the abundance of tar-O, which approaches 40% of the tar mass. This oxygen converts most of the tar into CO rather than soot during tar decomposition. Essentially all the fuel-N is released as NH3 during secondary pyrolysis. Note that the abundance of NH3 with switchgrass, and its higher soot yield and lower CO yield, are the major differences between the two biomass forms. In contrast, the secondary pyrolysis products from all the coals are dominated by soot which is the most abundant product, by far. The GHC yields are comparable from all coals but less than a fourth of those from the biomass. Hydrogen yields are also comparable and double those from biomass. The yields of the oxygenated gases diminish with coals of progressively higher rank, like coal-O levels. But even the highest CO yield from JR is only about one-quarter the CO yield from the biomass. The only predicted N-species is HCN although, in actuality, a minor amount of NH3 may have been released from JR coal (but none of the others). The N-species yields are directly proportional to the coal-N levels, as expected. The most extensive coverage of fuel quality impacts in the testing program used co-milled biomass injection. This focus is sharpened by considering all such data

228

Process Chemistry of Coal Utilization

500

400

600

300

500

400

200 3.5 % O2, 15 % OFA CO-Milled Injection

100 100

200

300

400

Measured NOx @3% dry O2

500

3.5 % O2, 0 % OFA CO-Milled Injection 200

300

400

500

600

700

300

Predicted NOx @ 3% dry O2

Predicted NOx @ 3% dry O2

700

800

Measured NOx @3% dry O2

Fig. 6.27 Parity plots for the data subset with co-milled injection, 3.5% O2, and (left) 15% OFA; and (right) 0% OFA, where the cases (□) with NH3 injection are distinguished from (●) all other test results.

at 3.5% O2 in separate groups for 15% and 0% OFA. Summary evaluations for these two subsets appear in Fig. 6.27. The criterion for evaluating the model predictions is the standard error of estimation (SSE), which is defined as follows: vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u nS uX p 2 u P i  Pi O u t i¼1 SSE ¼ nS  nF  1

(6.1)

where nS is the number of records under evaluation; Pi p is the prediction for the ith record; Pi O is the measured value; and nF is the number of independent factors accounted for in the model. The number of independent modeling factors is easiest to specify when the model is a multivariate regression; however, it is ambiguous when mechanistic models are involved in the predictions. For the evaluation of fuel quality impacts in this section, nF was specified as 4, because fuel quality impacts in the modeling were primarily expressed through the C, H, O, and N contents of the various fuels. As nS is much greater than nF, the specification on nF is unimportant. A parity plot for the NOX emissions for all cases with co-milled injection, 3.5% O2, and 15% OFA appears in Fig. 6.27. This data subset represents the complete range of fuel quality in the testing program. The range of measured values, from 140 to 460 ppm, reflects the range in coal quality from subbituminous through lv bituminous, compounded by the co-firing with both biomass forms. The predicted values show no systematic discrepancies over the entire range. The SSE is 32.4 ppm, which is roughly double the standard deviation assigned for the measured values. The r2-correlation coefficient for this data subset is 0.834, and the F-factor is 578. The predictions clearly depict the fuel quality impacts within useful quantitative tolerances over the full range in the testing program, without bias. This performance is especially significant because none of the fuel quality parameters in the modeling were adjusted once their

ChemNet furnace applications

229

values were specified for the baseline, coal-only cases. Only the proximate and ultimate analyses, grind size fractions, and biomass loadings were changed (as in the testing program) to achieve the agreement in Fig. 6.27. The parity plot for the NOX emissions for all cases with co-milled injection, 3.5% O2, and 0% OFA also appears in Fig. 6.27. As with the subset for 15% OFA, the range of measured values reflects the broad range in fuel quality in the testing program, but it has been shifted upward to 300 to 750 ppm by the higher near-burner SR in unstaged flames. The predicted values show no systematic discrepancies over the entire range. The SSE is 59.0 ppm, which is almost double the value for the 15% OFA subset. The r2-correlation coefficient for this data subset is 0.718, and the F-factor is 340. The statistics are weaker, in part, because of greater uncertainties in the tests with unstaged flames. Notwithstanding, the predictions clearly depict the fuel quality impacts within useful quantitative tolerances over the full range in the testing program, without bias. Only the proximate and ultimate analyses, grind size fractions, and biomass loadings were changed (as in the testing program) to achieve the agreement in Fig. 6.27. Another important aspect of the fuel quality impacts is the dependence on the biomass firing level. This aspect is apparent in the predictions for individual test series in Table 6.5. In 9 of 10 of the coal/biomass combinations with 15% OFA, the measured Table 6.5 Evaluation of predicted NOX for co-milled flames with 3.5% O2. NOX, ppm @ 3% dry O2 15% OFA

0% OFA

Series

Fuel

Pred

Meas

Pred

Meas

1

PR hv bit PR/10%SD PR/20%SD PR/15%SG PR/20%SG PR hv bit GL hv bit GL/10%SD GL/20%SD PR hv bit PR/20%SD GL hv bit GL/10%SG GL/20%SG JR subbit JR/10%SD JR/20%SD JR/10%SG JR/20%SG

325 255 236 277 258 354 372 365 318 351 314 368 399 349 263 221 212 203 177

328 271 245 269 260 346 360 273 327 319 270 356 340 336 240 191 189 185 142

464 365 363 363 384 453 453 459 434 453 411 449 501 456 395 332 302 305 369

447 433 444 413 367 465 472 468 476 496 434 483 477 525 398 358 313 333 318

5

6

7

Continued

230

Process Chemistry of Coal Utilization

Table 6.5 Continued NOX, ppm @ 3% dry O2 15% OFA

0% OFA

Series

Fuel

Pred

Meas

Pred

Meas

10

PR hv bit PR fine GL hv bit GL/5%SD GL/10%SD GL/20%SD JW lv bit JW/5%SD JW/10%SD JW/20%SD JW lv bit JW/5%SG JW/10%SG JW/20%SG

315 235 398 331 313 251 419 416 417 436 429 422 442 404

303 284 345 347 305 281 422 420 389 364 450 455 440 407

452 430 602 693 626 588 639 546 548 542 664 545 589 512

495 436 568 523 571 478 632 605 522 635 640 664 617 426

11

12

13

NOX emissions decreased for progressively higher biomass loadings. The only exception is for series 5 with SD on GL, for which the measured emissions pass through a minimum with biomass loading. The predictions exhibit the correct tendency for lower emissions with higher biomass loadings in 7 of 10 cases. This trend is illustrated in Fig. 6.28 for the PR-only baseline and the two PR/SD cases from test series 1. The case with PR-only is a calibration point, so the close agreement reflects parameter adjustments. But the predictions for the two cases with SD loadings of 10% and 20% demonstrate that the modeling capability for the impact of biomass loading is essentially within measurement uncertainty with no parameter adjustments at all. Among the three-test series for which the predicted impact of biomass loading is not monotonically decreasing, two are inaccurate. The predictions for series 13 for JW/SG show essentially no NOX reduction for SG loadings of 5% and 10%, in accord with the data, and a reduction to 404 ppm versus a measured value of 407 ppm for a loading of 20%. Whereas the predictions for GL/SD from series 6 are accurate for 0 and 20% SD, the one for 10% SG is 60 ppm too high. Similarly, the predictions for JW/SD from series 12 are accurate at 0 and 5% SD, but too high by up to 70 ppm for 10% and 20% SD. The effect of biomass loading is more complex in unstaged flames. Among the ten fuel combinations in this subset, only six had measured NOX emissions that monotonically decreased for progressively higher biomass loadings. Two combinations exhibited no change (PR/SD from series 1 and GL/SD from series 5); one exhibited a minimum (JW/SD from series 12); and one had higher emissions with the highest biomass loading (GL/SG from series 6). The predictions for this subset

ChemNet furnace applications

231

350

NOX@ 3% dry O2

325

300

275

250

225

200

hv bit/SD, 3.5 % O2, 15 %OFA Co-Milled Injection 0

5

10

15

20

Biomass Loading, wt. %

Fig. 6.28 Evaluation of the predicted impact of biomass loading on NOX emissions for hv bituminous/sawdust flames.

were also more varied. Three cases had substantial NOX reductions for 10% biomass but no further reductions for 20%. Three (PR/SD from series 6; JR/SD from series 7; and GL/SD from series 11) had monotonic decreases with higher loadings; two showed no change (GL/SD from series 5 and GL/SG from series 6); and one exhibited a minimum in NOX vs. loading (JR/SG from series 7). Evaluations for three of these dependences appear in Fig. 6.29. The evaluation from series 13 illustrates a substantial predicted reduction for low loadings, but minimal additional reduction through 15% biomass, whereas the data show minimal change through 10% SG but much lower NOX with 20%; that from series 5 shows no change; and that from series 7 shows a monotonic decrease. The predictions are essentially within experimental uncertainty over the full range of biomass loadings, except for the highest loading in series 13. The detailed evaluations for all other cases in this data subset are collected in Table 6.5. Staged flames produce less NOX than unstaged flames, and this behavior is clearly evident in the predictions in Table 6.5. In addition to staging, the testing program characterized variations in furnace stoichiometry and, to a much lesser extent, variations in particle size. Furnace stoichiometry was varied to obtain flue gas O2 levels from 2.5% to almost 5%. The predictions over this range in the various test series with 15% OFA are evaluated in Fig. 6.30. For each series, predictions for a coal-only baseline are evaluated along with those for one or two co-fired cases. In series 1 and 7, the coals were cofired with both biomass forms, whereas the rest were cofired with either SD or SG. All cofiring cases in Fig. 6.30 had 20% biomass. The error bars indicate the standard deviations reported by the testing team for each run.

232

Process Chemistry of Coal Utilization

700 3.5 % O2, 0 %OFA Co-Milled Injection

NOX@ 3% dry O2

600 JW/SG Ser. 13 500 GL/SD Ser. 5 400

JR/SD Ser. 7

300 0

5

10

15

20

25

Biomass Loading, wt. %

Fig. 6.29 Evaluation of the predicted effect of biomass loading on NOX emissions for test series with co-milled injection, 3.5% O2, and 0% OFA with (JR) subbituminous, (GL) hv bituminous, and (JW) mv bituminous coals.

The predictions are generally within measurement uncertainties across the full range of flue gas O2 for all coal types and all cofiring cases. The only sizeable discrepancies are for the PR-baseline in series 1 for O2 levels below 2.7%; for the JR/20%SG case in series 7 for O2 levels below 3%; and for the JW-baseline in series 13 for O2 levels below 3%. Cofiring (up to 20 wt%) hardly perturbs the measured dependence on furnace stoichiometry, and this tendency is evident in the predictions for all cases except series 6. But even this distinctive O2 dependence remains within the measurement uncertainties.

6.3.2.5 Outlook for ChemNet simulations of pilot-scale coal flames The ChemNet analysis of the CRF furnace establishes several milestones in simulations of large-scale, commercially relevant pc flames, as follows: (1) For the first time, the complete range of fuel quality in biomass cofiring applications was simulated without heuristic parameter adjustments. None of the parameters in the mechanisms for devolatilization, soot chemistry, or chemistry in the gas phase, including N-species conversion, were adjusted to improve fits with measured NOX levels in the flue gas. The conversion factors for char-N and the initial char oxidation reactivities were specified to match the NOX emissions from coal-only flames with 15% OFA and 3.5% flue gas O2. These values were not changed to simulate any of the other fuel combinations or operating conditions. (2) Only a handful of CFD simulations were required to enable predictions with the complete reaction mechanisms over the full operating domain. The same extrapolation procedures for

ChemNet furnace applications

233

Series 1, Pr/SD/SG, 15% OFA CO-Milled Injection

NOx@ 3% dry O2

500

400

300

200

Series 5, GL/SD, 15% OFA CO-Milled Injection

Series 7, JR/SD/SG, 15% OFA CO-Milled Injection

NOx@ 3% dry O2

500

400

300

200

Series 6, GL/SG, 15% OFA CO-Milled Injection

NOx@ 3% dry O2

500

400

300

200

Series 13, JW/SG, 15% OFA CO-Milled Injection

Series 12, JW/SG, 15% OFA CO-Milled Injection 2.0

2.5

3.0

3.5

4.0

Exit O2, Wet %

4.5

5.0

5.5 2.0

2.5

3.0

3.5

4.0

4.5

5.0

5.5

Exit O2, Wet %

Fig. 6.30 Evaluation for the full range of flue gas O2 with 15% OFA and co-milled injection with (JR) subbituminous, (PR, GL) hv bituminous, and (JW) mv bituminous coals. Each case contains a coal-only baseline (4, dashed curve) and, if available, cofired cases with (●, solid curve) 20% sawdust (SD) and (, dotted curve) 20% switchgrass (SG). staging and furnace stoichiometry were applied to the entire data subset. ChemNet incorporates additional information from CFD simulations but does not entail any additional computational burden. Each CRF flame simulation took 15 to 50 min on a 1.5 GHz microprocessor. (3) ChemNet accurately described how cofiring with biomass that has both much less and much more fuel-N than the coal component will reduce NOX emissions. The rudimentary NOX production submodels in CFD cannot predict less NOX when any blend component contains more nitrogen than the baseline coal. The chemistry underlying volatile-N

234

Process Chemistry of Coal Utilization

conversion in NBFZs governs NOX emissions, which are not necessarily proportional to the N-contents of the fuels. (4) Predicted NOX emissions for co-milled injection, 15% OFA, and 3.5% O2 represent the complete range of fuel quality within useful quantitative tolerances. The SSE for this evaluation is 32.4 ppm, which is twice the stated experimental uncertainty but within the scatter in the data for repeated cases. These predictions show no systematic discrepancies over the entire range of fuel quality. The predictions also depicted the dependence on biomass loading within useful quantitative tolerances for all fuel combinations. Predicted NOX emissions from unstaged flames also show no systematic discrepancies over the entire range of fuel quality. The SSE for this evaluation is 59.0 ppm, reflecting larger measurement uncertainties as well as extrapolations away from the staged flames that were characterized with CFD simulations. (5) Predicted NOX emissions over the full range of furnace stoichiometry were generally within measurement uncertainties for both staged and unstaged flames, which validates the extrapolation procedure developed for this operating condition.

The unparallel quantitative accuracy of the predicted NOX emissions can be attributed to ChemNet’s fine resolution of chemistry in near-burner flame regions. Indeed, the predicted flue gas NOX emissions can be resolved into separate contributions from near-burner NO; NO reduction in flame cores by HCN and NH3; conversion of residual HCN and NH3 at the inlet to the mixing layer; and the inevitable conversion of fixed portions of char-N into NO in all char burnout regions. The regional operating conditions assigned directly from the CFD simulations gauge these various contributions. The structures of the coal-only and the biomass cofired flames in the CRF are qualitatively similar: All exhibit a very rapid surge in the NO level immediately after injection, due to the ignition of volatiles under the very lean conditions associated with primary coal streams during the initial stages of devolatilization. But as more volatiles are released, the gas phase becomes progressively more reducing, which enables the early NO to be reduced into HCN and NH3. The extent of reduction is determined by the SR-value among the gaseous species only and the residence time available before the core fluid is exposed to secondary air in the mixing layer. All fixed-N species are rapidly converted into NO and N2 at the beginning of the mixing layer, which then sustains the oxidation of soot and char. From this point onward, chemistry in the gas phase is inconsequential while the conversion of some char-N into NO supplements the NOX inventory. The NO emissions from CRF flame cores are governed by the proportions of HCN and NO at the end of devolatilization, provided that the residence time is sufficient to complete the NO reduction. This factor alone is responsible for the significantly higher core-NO with one of the hv bituminous coals (GL) than the other (PR), and it is the dominant factor underlying the excessive NO emissions from cores with lv bituminous. An overabundance of HCN in cores with the subbituminous is responsible for its very low core-NO level, which could have been even lower given additional residence time in the flame core. The only universal effect of biomass cofiring on NOX emissions is the reduction in the amounts of char-N released into the char burnout regions. Cofiring can conceivably

ChemNet furnace applications

235

perturb volatile-N conversion by promoting prompt NO formation or, conversely, by reducing the maximum NO level in an NBFZ. But the simulations show that the main effect on volatile-N conversion is to shorten the core residence time, which conveys more NO, HCN, and NH3 into the early mixing layer, where it may or may not lower the NO contribution from volatiles combustion. The various alternative routes are determined by the chemistry among the biomass form and the baseline coal, and cannot be foreseen from the standard fuel properties. Two factors are primarily responsible for the significant NOX reduction observed for CRF flames with biomass cofiring. First, there is more volatile matter from biomass, and it contains much less soot. Consequently, the near-burner SR-values for the gas phase are significantly richer than in coal-only flames. Under richer conditions, more near-burner NO will be reduced away, and a greater proportion of the fixedN species will be converted into N2. CRF flames provide sufficient residence times in the flame cores for NO reduction and fixed-N conversion to N2. Second, a significantly lower percentage of the total fuel-N will remain in the char beyond the NBFZ, where its partial conversion to NO is inevitable. If the biomass contains nitrogen, it releases all of it in the NBFZ. If the biomass contains no nitrogen, then the inventory of char-N beyond the flame core is reduced in proportion to the biomass loading. In either case, there is less char-N beyond the point where NO can be reduced by chemistry in the gas phase. The first factor is especially sensitive to fuel quality. Indeed, even the NOX emissions from the four coal-only flames are ordered according to the volatiles yields and contributions from soot. While the volatiles yields decrease and the soot contributions increase for coals JR, PR, GL, and JW, the baseline NOX emissions in the staged flames increased from 240 to 338 to 358 to 436 ppm (cf. Table 6.5). Biomass cofiring mitigated these differences by enhancing volatiles yields and by reducing the soot contributions. The net effect was significant NOX reduction for the cofired flames, even when the biomass supplements the inventory of fuel-N. The extent of reduction was directly proportional to the biomass loading with both forms of biomass on subbituminous and PR hv bituminous. But it was only proportional to the switchgrass loading in flames cofired with low volatility coal; cofiring lv bituminous with sawdust was ineffective at all but the highest loadings. Cofiring GL hv bituminous reduced NOX, but the magnitude was disproportionate at 10% loadings with both biomass forms. Whereas unstaged CRF flames had only slightly more O2 in their flame cores than staged flames, even small increases in the O2-inventory will significantly shift the N-speciation toward more NO and less HCN in the vicinity of the stoichiometric point. Consequently, the extents of NO reduction in the cores of unstaged flames are significantly lower than in staged flames, and all HCN and NH3 were eliminated in the flame cores. Less char-N is carried into the mixing layer than in staged flames, simply because unstaged flames sustain higher extents of char burnout due to the greater availability of O2. But the flue gas NOX emissions are greater, due to the less effective reduction of near-burner NO in the flame cores. Similarly, the main effect on volatile-N conversion at progressively higher O2 levels is to decrease the inventory of HCN available for near-burner NO reduction.

236

Process Chemistry of Coal Utilization

Consequently, more NO and less HCN and NH3 leave the flame core at higher O2 levels, and the NO produced from volatiles is correspondingly higher. This detrimental effect is partially compensated for by lower values of char-N, but the predicted flue gas NOX emissions are still greater for higher flue gas O2 levels, as they should be.

6.3.3 PC combustion at commercial-scale This section presents case studies for ChemNet applications on full-scale utility furnaces. As these applications are inevitably labor-intensive and sometimes excessively so, they should only be developed from a clear understanding of why an analysis based on comprehensive chemistry is warranted, and which specific information that analysis will deliver. Presumably, other simpler calculations would have been performed in early attempts to obtain the required information but had been found lacking in scope, execution, or precision. Only then should a ChemNet analysis be evaluated more formally. Given a decision to proceed, the next issue is whether the subject of the analysis is confined to only a portion of the furnace flowfield. If so, then ChemNet should only be applied to that particular portion to tightly focus the scope of the analysis. Fig. 1.3 in Chapter 1 of the first volume of this series (reproduced below in Section 6.3.3.2) presented mean thermal histories through several portions of commercial furnaces with and without staging. The temperature differences among some regions can differ by several hundred degrees, which is definitely large enough to localize certain types of chemistry. For example, the levels of VOCs in quench layers along furnace waterwalls significantly affect VOC stack emissions because these compounds are rapidly burned away in hotter furnace regions. The same pertains to CO except that CO levels at a furnace exit also reflect contributions from the CO from char burnout along convective passes, where temperatures may be too cool to completely oxidize CO into CO2. Conversely, the greatest contributions to furnace NO levels from oxidation of the N2 in the air—so-called thermal NO—are made along the furnace core flow, where temperatures are hottest. Furnace temperature profiles, either measured or from CFD, should be intersected with the subject of the analysis to target only as much of a furnace as necessary. The added benefit of such a preliminary evaluation is that some portions of a furnace flowfield are predominantly in turbulent plug flow, for which a suitable equivalent reactor network is simply an extended CSTR-series. No formal tracking analysis of CFD flowfields is required when the flow and temperature fields can be translated into network specifications by inspection. The analysis of the pilot-scale CRF furnace in the previous section showed that NBFZs sustain the most complex chemistry, by far, where the reaction system proceeds through several stages of both homogeneous and heterogeneous chemistry with numerous fuel components. In contrast, the chemistry in OFA injection zones is governed by mixing intensities, with only a minor role for homogeneous chemistry in the equilibration of water-gas shifting. Burnout zones support an even smaller role because of the cooler temperatures. When the targets of an analysis are confined to the upper furnace elevations, it is not always necessary to analyze the NBFZ in detail to specify the gas composition at the inlet to the furnace

ChemNet furnace applications

237

region of interest. The NBFZ analysis can be circumvented with much simpler calculations including, among many others, thermochemical equilibrium calculations. Then, the reactor network would contain distinct branches for close-coupled and separated OFA injection zones and for the burnout zone, which may or may not be entraining fluid from the quench layers. However, some targets of an analysis such as furnace NOX and LOI emissions, simply require a full and formal ChemNet CFD postprocessing analysis. Full analyses are also necessary if the goal is to compare the chemical flame structures among different firing configurations, or among different combinations of NOX control technologies. ChemNet also provides the most secure technical foundations for simpler calculations to be used in repeated analyses of the CQ impacts. These approaches are illustrated in two succeeding case studies on portions of a furnace and one full furnace analysis.

6.3.3.1 The impact of heat absorption on furnace NOX levels This case study pertains to the chemistry responsible for greater NOX emissions from a 560 MW furnace when the heat absorption in the lower furnace elevations was impeded by slag deposits along the NBFZ. The furnace is opposed-wall fired through 30 low-NOX burners with CCOFA immediately above the burner belt and SOFA injectors 6.5 m further up the furnace. The furnace is fired with a standard utility grind of hv bituminous coal. Temperature profiles near the centerline were estimated from CFD for two heat absorption profiles, one for baseline operation and one for absorption impeded by slag deposits up to the CCOFA elevation. As seen in Fig. 6.31, the mean gas thermal histories along the furnace centerline depend on the heat absorption rate, especially during the period from 0.25 to 1.5 s. During this interval, the history

Fig. 6.31 Mean gas thermal histories from CFD for (dashed curve) baseline and (solid curve) low absorption conditions.

238

Process Chemistry of Coal Utilization

for the low absorption case is hotter by up to 130°C than the baseline. For longer times the histories are the same. The apparent impact of impeded absorption on NOX emissions was huge as the emissions at the centerline of the furnace exit increased from 130 to 210 ppm dry @ 6% O2. A ChemNet analysis was developed to determine whether or not the differences between the thermal histories in Fig. 6.31 could be responsible for the greater NOX emissions with impeded heat absorption, all else the same. Its scope was restricted to the furnace elevations beyond the NBFZ because of the deep staging across the NBFZ. As percentages of the total furnace air flow, the primary and secondary air into the burner belt account for 21% and 44%, respectively, whereas CCOFA and SOFA account for 11% and 24%. The furnace air staging ratio is 35%. As the formal ChemNet analysis for this firing configuration is so complex, an abridged analysis was developed by analyzing a portion of the NBFZ with an equilibrium gas composition instead of the full kinetic analysis. The thermochemical equilibrium calculation was intended to specify a gas composition into a reactor network for the furnace core flow beyond most of the NBFZ. The equilibrium gas composition would represent coal conversion to the point where all GHCs and soot had been converted and all near-burner NO had been reduced away. In terms of the flame structure for a CRF pilot-scale flame of hv bituminous coal in Fig. 6.20, this point corresponds with the structure from 200 to 300 ms, when gas temperatures are hottest; all GHCs have burned away; O2 levels are low and CO levels are appreciable because secondary air has not yet fully mixed into the flow; and most of the soot and much of the char have burned out. This system was simplified further by basing the equilibrium gas composition on the complete distribution of secondary pyrolysis products, including soot. For this particular coal, the predicted soot yield was nearly 42 daf wt% which is actually greater than the yield of combustibles in char. Consequently, the equilibrium calculation omitted char entirely and consumed all soot as a means to account for the incomplete and simultaneous conversion of char and soot in the NBFZ. An O2 level for the equilibrium calculation was set to give equal mass fractions of CO and CO2 in the equilibrated gas, as a means to ensure that the equilibrium gas composition included appreciable CO and H2. The estimated gas composition from the NBFZ contained 13% by mass of CO and CO2 plus half that level of moisture but no GHCs, as intended. The network for the furnace NOX calculations appears in Fig. 6.32. It comprises the equilibrium calculations for the first portion of the NBFZ, then CSTR-series for the remainder of the NBFZ (denoted as the Post-Flame (PF) Zone), the CCOFA zone, and SOFA zone. No volatiles or tar decomposition products are injected into any of the CSTRs because these fuel components were burned away in the equilibrium calculation. All succeeding CSTRs account for simultaneous homogeneous combustion, char burnout, and N-species transformations. Preliminary calculations indicated that the predictions were very insensitive to the numbers of CSTRs in each zone, so these parameters were assigned as the minimum numbers that could adequately represent variations in the temperature profiles across each zone. However, the mixing time constants for secondary air, CCOFA, and SOFA strongly affected the predicted NOX levels at the furnace exit, as expected. These

ChemNet furnace applications

PRIMARY FLAME ZONE

239

POST-FLAME ZONE

Coal Primary Air

CCOFA Air Secondary Air

CCOFA AIR INJECTION ZONE OFA Air OFA AIR INJECTION ZONE UBC+flyash Exhaust Gases

Fig. 6.32 Reactor network for the core flow along the furnace.

constants were adjusted to match the predicted and reported NOX emissions for the baseline conditions. The values agreed only if mixing rates were as slow as possible in the first two zones, then fairly rapid to complete char oxidation in the SOFA zone. The best results in the baseline case were obtained when half the secondary air was carried over to the PF zone; 40% of the CCOFA was carried over to the SOFA zone; and nearly all air was mixed into the flue gas before the furnace exit. With these mixing levels, the predicted NOX in flue gas is 137 ppm @ 6% O2 vs. the reported value of 130 ppm. Similarly, the char oxidation reactivity in CBK/E was adjusted until the predicted UBC for baseline conditions of 1.5 wt% matched the measured values of 1.3% to 1.5 wt%. All operating conditions and parameter values were kept the same when the low absorption case was simulated. The only difference was that the hotter thermal history in Fig. 6.31 was incorporated into the ChemNet calculations. Various aspects of the results are reported in Fig. 6.33 for baseline conditions and the low absorption case. In these figures, the results are plotted versus the index on each CSTR in the network. On the X-axis, the PF-zone comprises the first four CSTRs; the CCOFA-zone comprises the next five; and the SOFA-zone comprises the final six. For baseline conditions, the SR-values increase from 0.7 to 0.9 over the PF-zone, from 0.9 to just under unity in the CCOFA-zone, and from unity to 1.15 in the SOFAzone, while O2 levels increase across each zone. Oxygen remains under 100 ppm across the PF-zone, then increases to approximately 500 ppm in the CCOFA-zone. The gradients in the O2 level are greatest across the SOFA-zone, coincident with the surging degrees of burnout. Ultimately, the flue gas O2 level approaches 3%, as it should. Less than 5% of the char is burned away in the PF-zone, which indicates that the high concentrations of gaseous fuels leaving the PF-zone effectively compete

240

Process Chemistry of Coal Utilization

Fig. 6.33 Furnace performance characteristics versus the index for CSTRs in the reactor network for the (left) baseline and (right) low heat absorption/high-temperature cases. The results include (top) temperature and SR for the gas phase; (middle) O2 mole fraction and percentage of char burnout; and (bottom) normalized NO production rate and NOX concentration.

for O2, as expected. The competition advances toward greater char oxidation in the CCOFA-zone, where one-quarter of the char is consumed, but most of the char burns out in the SOFA-zone, with almost half consumed in the first CSTR. Burning rates in succeeding CSTRs fall continuously even while the O2 level increases, due to thermal annealing of the char. The lower panel in Fig. 6.33 shows the NOX production rate in each CSTR normalized by the rate in the first CSTR, and the (uncorrected) NO concentrations leaving each CSTR. Most of the NO is produced in the PF-zone, despite the very low O2 levels. Even though NO production rates fall continuously across every zone, more than half the maximum NO concentration is reached in the first two CSTRs in the PFzone. There is a similar surge across the first two CSTRs in the CCOFA-zone. Then

ChemNet furnace applications

241

the NO levels approach a maximum value of 218 ppm before relaxing to lower values across the OFA-zone, primarily due to the high levels of dilution as the SOFA stream mixes with the flue gas. The analogous results for the low heat absorption case show significantly hotter gas temperatures than the baseline values through the first seven CSTRs, but the same thereafter. The profiles of SR-values for the gas phase and O2 are nearly the same as the baseline’s across the entire network. The extents of char burnout are slightly greater across the first six CSTRs due to the hotter combustion temperatures, then they relax to very similar values during the later stages. The ultimate predicted UBC value is 2.1 wt%, which is greater than the baseline value due to the more extensive thermal annealing at the hotter temperatures in the low absorption case. The normalized NO production rate is nearly the same in the low absorption case as it is in the baseline case. Yet the NOX levels are much greater across the entire reactor network. They surge to 300 ppm across the PF-zone then continue to approach a maximum value just under 350 ppm at the inlet to the CCOFA-zone. After the second CSTR in the CCOFA-zone the NOX level relaxes continuously, reaching a level of 236 ppm in the flue gas. This value is only 5% less than the reported raw value of 259 ppm. The corrected value of 194 ppm @ 6% O2 compares similarly well with the reported value of 210 ppm. Even though the profiles of SR, O2 level, char burnout, and normalized NO production rate for the low absorption case are virtually identical to those for the baseline case, the simulations give significantly greater flue gas NOX concentrations, in good quantitative agreement with the measurements. This agreement was obtained without any parameter adjustments whatsoever. The only difference in operating conditions between the two simulation cases is the temperature profiles. All other conditions and model parameters were the same. Sensitivity analysis on the homogeneous chemistry identified the specific reasons for the enhanced NO levels. The main channels for NO production and destruction appear in Fig. 6.34 for the baseline case. The contribution for each channel is expressed as a fraction of the total net production or destruction rates in each CSTR. They do not exactly sum to unity because several minor channels have been omitted from each stage. Both the production and destruction pathways are surprisingly simple, and there is a continuous mechanistic shift across the three zones of the furnace. Throughout the PF-zone, the dominant NO production channel is HNO + ðH or OHÞ ! NO + ðH2 orH2 OÞ It is not surprising that the destruction of HNO by OH becomes more important as the O2 level increases across the PF- and CCOFA-zones. Thermal NOX makes an appreciable contribution from two of the three channels in the extended Zeldovitch mechanism: N2 + O ! NO + N N + OH ! NO + H

242

Process Chemistry of Coal Utilization

1.0

Fraction of Total Production

Base Case 0.8 NO2+OH 0.6

HNO+H

0.4 HNO+OH 0.2

NO2+O N+OH N2+O

NO2+H

NO+H+M

NO+O+M

0.0

Fraction of Total Destruction

0.8

0.6 Base Case 0.4

0.2 NO+H+N2 0.0

0

2

4

6

8

NO+OH+M 10

12

14

16

18

Reactor Number

Fig. 6.34 Major reaction channels for NO (top) production and (bottom) destruction for the baseline case versus the CSTR index, where curves are labeled by the reactants in the channels.

As seen in Fig. 6.34, each of these four channels diminishes across the PF- and CCOFA-zones, becoming negligible at the inlet to the SOFA-zone, where they are supplanted by NO production through the NO2 intermediate: NO2 + ðH or O or OHÞ ! NO + ðH2 or H2 OÞ As expected, the primary radical chain carrier shifts from H to OH to O as the O2 level progressively increases across the furnace. There are even fewer destruction channels in Fig. 6.34 to consider. The attack of NO by H dominates across the PF- and CCOFA-zones, at which point O becomes the primary radical agent with minor contributions from OH.

ChemNet furnace applications

243

0.6

0.6 Base Case

High-T Case 0.5

HNO+H 0.4

0.4 HNO+H HNO+OH

0.3

0.3 HNO+OH

0.2

0.2 NO2+H

NO2+H 0.1

0.0

N+OH N2+O

N+OH N2+O 1

2

3

4

1

2

Reactor Number

3

Fraction of Total Production

Fraction of Total Production

0.5

0.1

0.0

4

Reactor Number

Fig. 6.35 Major channels for NO production for (left) baseline and (right) low absorption cases vs. the CSTR index in the primary flame zone, where curves are labeled by the reactants in the channels.

The relative contributions for these major channels for the low absorption case are compared to the baseline behavior in Fig. 6.35. Only the channels for the PF-zone appear because the major channels for both other downstream zones are virtually identical. The contributions from NO2 + H are the same for both cases, and that from HNO + OH is only slightly higher for the baseline case. The primary difference is a significantly greater contribution from thermal NOX, especially in the third and fourth CSTRs in the PF-zone. As is necessary to satisfy the definition of the fractional contributions, there is a corresponding reduction in the contribution from HNO + H for the low absorption case. However, the shift toward greater contributions from thermal NOX channels does not mean that most of the additional NOX from the low absorption case is, in fact, thermal NOX. The reason is apparent in Table 6.6, which presents the actual NO production rates from the three major channels in Figs. 6.34 and 6.35. In both the baseline and low absorption cases, NO production via an HNO intermediate proceeds three to four times faster than thermal NO production. The margin is even greater for the last two CSTRs in the baseline case. In turn, thermal NO production is significantly faster

Table 6.6 NO production rates from three major channels, in 109 mol/cm3-s. CSTR

1 2 3 4

NO via HNO

Thermal NO

NO via NO2

Base

Low abs.

Base

Low abs.

Base

Low abs.

5.19 6.30 5.70 4.19

6.49 9.09 4.68 3.43

1.38 1.52 0.44 0.31

2.15 2.53 1.82 0.86

0.30 0.80 0.90 0.94

0.46 1.52 1.03 0.95

244

Process Chemistry of Coal Utilization

than the production of NO via an NO2 intermediate in the first two CSTRs; in the last two CSTRs, these rates become comparable in the low absorption case and are reversed for the baseline case. As an approximate estimate for the relative contributions to the total NO concentrations from the PF-zones in the two cases, the NO was apportioned according to the magnitudes of the NO production rates. This exercise suggests that about 16% of the NO leaving the PF-zone for the baseline case is thermal NO, whereas the percentage increases to 21% for the low absorption case. In terms of the difference in the NO concentrations leaving the PF-zones from both cases, which is 143 ppm, enhanced thermal NO accounts for 25%, and faster production rates for NO via the HNO intermediate account for slightly more than 60%. The first-order, normalized logarithmic sensitivity coefficients from the kinetics solutions identify which steps in the entire reaction kinetics exerted the greatest influence on NO production. Increasing the rates in the following three channels would have the greatest positive effect on NO production across the PF-zone in both the baseline and low absorption cases: N2 + O ! NO + N N2 O + H ! NO + NH N + OH ! NO + H Conversely, increasing the rates in the following four channels would have the greatest negative effect on NO production: H + OH + M ! H2 O + M ðwhere M is any third bodyÞ CO + O + M ! CO2 + M 2H + H2 O ! H2 + H2 O 2H + M ! H2 + M Considering that all the NO production channels involve species from the oxyhydroxyl radical pool, it is not surprising that radical recombinations inhibit NO production under furnace conditions. Notwithstanding some drastic simplifications for the NBFZ, the ChemNet simulations offer a compelling explanation for enhanced NOX emissions under low furnace heat absorption conditions. The three major channels for NO production beyond the NBFZ are radical decomposition of HNO, thermal NOX, and radical decomposition of NO2. HNO decomposition dominates across the PF- and CCOFA-zones, whereas thermal NOX is a significant contribution in only the PF-zone. NO production through an NO2 intermediate becomes progressively more important in the cooler SOFA-zone, where O2 levels are elevated. Low furnace heat absorption does not introduce any new major NO production channels, or even grossly distort the relative impact of the three major channels for baseline conditions. It simply increases all three NO production rates across the PF-zone and hardly affects any aspect of the kinetics in the CCOFA- and SOFA-zones. The rate of thermal NO production is enhanced the most so that its contribution to the

ChemNet furnace applications

245

total NO emissions increased from 16% for the baseline case to 21% for the low absorption case. However, about 60% of the total enhancement to NO emissions is due to a faster rate of radical decomposition of HNO. According to the simulations, most of the NO forms in the PF- and CCOFA-zones, especially the inlet regions to both zones where O2 levels are relatively high.

6.3.3.2 Paths to VOC and PAH emissions VOCs comprise very complex mixtures of saturated and unsaturated GHCs plus various oxygenates plus PAH. In some situations, mixture compositions are dominated by a handful of species that reflect the fuel being processed and the processing conditions. Generally speaking, VOC emissions from coal-fired furnaces are well below regulatory thresholds unless the fuel is badly maldistributed into a burner belt, or OFA mixing intensities are too weak, or the flowfield contains short-circuits through streams that are much cooler than the bulk flow through a furnace core. These connections have been revealed indirectly through field characterization work but had not yet been formally interpreted until recently. This case study uses ChemNet to identify which pathways through furnace flowfields produce and preserve VOCs and PAH into a furnace exit, and how coal quality affects these emissions. It is a component in a larger analysis that also covered VOC destruction along utility gas cleaning systems, as a means to identify surrogates for VOC and PAH stack emissions. However, only the furnace analysis is discussed here. The strategy is to survey VOC/PAH production and destruction along distinctive streams through furnaces with different firing configurations. The representative histories of furnace conditions in Fig. 6.36 were developed from CFD for three pathways through a full-scale T-fired furnace and one pathway each through staged and unstaged wall-fired furnaces. In the T-fired furnace, one path moved most of the reacting fuel particles into a cylindrical furnace core that rose upward along the centerline. Due to their substantial inertia, the fuel particles were able to penetrate the swirling tangential flow along the furnace walls and to entrain small portions of the primary and secondary air streams. The core expanded outward due to the addition of fuel suspension from different coal injection elevations, and to SOFA entrainment along the upper furnace elevations. This pathway features relatively slow heating to a maximum temperature of 1600°C and an abrupt consumption of the available O2 within the first 250 ms. As the particle streams entrain only small portions of the injected air, the furnace core remains depleted of O2 throughout the lower furnace elevation. Oxygen from the SOFA jets eventually mixes into the furnace core while it moves through the convective passes. Whereas the furnace core flow moves radially inward and upward along the centerline, the second primary flow path swirls in a helical flow along the furnace walls. These so-called “gas ribbons” carry few fuel particles but have most of the volatiles released via devolatilization during the initial stages of particle heating. These volatiles rapidly mix with air from the primary fuel streams and CCOFA injectors. As seen in Fig. 6.36, the maximum temperatures are comparable to those in the furnace core

246

Process Chemistry of Coal Utilization

250

0 7

0 1750

6

1500

Temperature, °C

TRAD

TGAS

5

1000

4 O2

750

3

500

Temperature, °C

1

O2, Vol. %

2

250

1250

500

0 7

0

1500

6 O2

TGAS 4

750

TRAD 3

500

2 Quench Layer

0

Staged Wall-Fired

1 0 7 6

TRAD

TGAS

5 4 3

O2

2

Unstaged Wall-Fired 0.0

0.5

1.0

1.5

2.0 Time, s

2.5

3.0

3.5

1 0

5

1000

250

2

750

2 1

1250

3 O2

1000

0 1750

Gas Ribbons

4

1250

250

250

5

750 500

1500

TGAS

1000

500 Furnace Core

TRAD

O2, Vol. %

O2

Temperature, °C

3

O2, Vol. %

4

750

6

1250

O2, Vol. %

Temperature, °C

5

1000

7

1500

6

TRAD

1250

0 1750

Temperature, °C

1750

7 TGAS

1500

O2, Vol. %

1750

1 0

0.0

0.5

1.0

1.5

2.0 Time, s

2.5

3.0

3.5

Fig. 6.36 Histories of (solid curves) gas temperature, (dashed curves) radiation temperature, and (dotted curves) O2 mole fraction (left panels) through a T-fired furnace through the (top) furnace core, (middle) gas ribbons, and (bottom) quench layer; and (right panels) for central paths through (top) staged and (bottom) unstaged wall-fired furnaces.

but heating rates are faster, and the O2 levels rise much faster and reach much greater values. The third pathway is a quench layer that moves along the furnace walls. Whereas the quench layer in Fig. 6.36 happened to be compiled for T-firing, quench layers along waterwalls form under any firing configuration. A quench layer has the greatest O2 levels of all because it is the first portion of the furnace gases to contact SOFA, but also has significantly cooler temperatures than both other flowpaths. The two cases in Fig. 6.36 for staged and unstaged wall-firing are the expected conditions along the centerlines of flows from each burner. Both cases have faster heating to hotter maximum temperatures than the T-fired cases. And both have O2 concentrations that plummet very close to the burners and remain low for at least a second. The O2 level in the unstaged case never reaches the minimum in the staged case and gradually rises after about 500 ms, even while the gas temperature is as hot as

ChemNet furnace applications

247

1750°C. In the staged case, the minimum O2 level is maintained through 1.5 s, and the O2 level rises faster due to the direct injection of SOFA. These five combustion histories are implemented in a simulation series to determine (1) whether or not variations in firing configuration significantly affect VOC concentrations; and (2) the thermal history that maximizes VOC production. To resolve the various VOCs and PAH of interest, the elementary reaction mechanism for homogeneous chemistry in Chapter 5 was changed to one developed to describe the production of PAH from much lighter hydrocarbon precursors under flame conditions (Skjoth-Rasmussen et al., 2004). It consists of 773 reactions among 159 species. The mechanism emphasizes a broad range of substituted hydrocarbons with one and two rings, including benzene, toluene, and xylene (BTX) and various naphthalenes. Moreover, it describes the growth of various PAH with three and four condensed rings through pyrene, the most highly condensed PAH species with four rings. All the noncondensables from tar decomposition, secondary volatiles pyrolysis, char oxidation, and soot conversion can be incorporated into the gas phase mechanism as initial conditions and time-dependent injection rates into the reactors in a CSTRseries. This case uses the product distributions from FLASHCHAIN® for instantaneous secondary pyrolysis, with attenuations for residual PAH described below. The raw distributions are compiled in Table 6.7 for PRB subbituminous and two hv bituminous, Ill. #6 and Pit. #8. As expected, these distributions are dominated by soot and the oxygenated gases for all three coals. Whereas the yields of soot and H2 from all coals are comparable, total weight loss and the yields of CO, CO2, H2O, HCN, CH4, and C2H2 are greatest with the subbituminous. The Pit. #8 releases somewhat more soot and comparable amounts of GHCs, and the Ill. #6 releases the least amounts of GHCs and combustible gaseous volatiles, by far. Char oxidation was then evaluated over all the thermal and O2 histories in Fig. 6.36. Fig. 6.37 illustrates this procedure with the predicted particle temperature and extents of conversion for char and soot for 161 μm particles of Pit. #8. Although the X-axis begins at 0 s, the time coordinate for char conversion actually starts 250 ms into the furnace residence time, once devolatilization has finished. The particle temperature rises from 200 to 1500°C in 260 ms. Char conversion is very slow through the first 2 s due to the very low O2 levels, then increases rapidly thereafter as the O2 level rises via air entrainment. The same form is apparent in the extent of soot conversion, albeit at a faster rate due to soot’s much smaller particle size. The respective ultimate conversions for char and soot are 87% and 100%. However, for a mean size of 55 μm in Table 6.7 Predicted secondary pyrolysis products from three coals, in daf wt%. Fuel

Char

Vol.

Soot

CH4

C2H4

H2

CO

CO2

H2O

HCN

H2S

PRB Ill. #6 Pit. #8

39.9 49.2 49.9

60.1 50.8 50.1

25.4 25.5 29.4

0.69 0.31 0.41

1.80 0.00 1.37

3.82 3.78 3.95

12.7 7.5 6.0

6.6 3.9 2.0

7.9 6.8 4.7

0.77 0.68 0.61

0.53 2.33 1.7

248

Process Chemistry of Coal Utilization

100

1600

Temp

1200 1000

60

800 40

600

Soot

20

400

Char Staged Wall-Fired

0

Particle Temperature, C

Burnout of Char & Soot, %

1400 80

200 0

0.0

0.5

1.0

1.5 2.0 Time, s

2.5

3.0

Fig. 6.37 Predicted extents of conversion for (solid curve) char and (dotted curve) soot and (dashed curve) a particle temperature history for 161 μm particles of hv bituminous under the staged wall-fired conditions in Fig. 6.36.

the standard p. f. grind, char burnout was complete for all three coals with all five sets of furnace conditions. To corroborate the conditions assigned for the five furnace pathways, LOI was estimated with the Ill. #6 and Pit. #8 samples with char oxidation simulations for five size cuts from 161 to 297 μm; smaller size cuts were completely converted. The contributions from each size cut were weighted by the mass fractions for these sizes in a conventional p. f. grind with 70% through 200 mesh (74 μm) and 0.5% on 50 mesh (297 μm). The subbituminous was omitted because its combustion efficiency is much greater than both other coals, as expected. As seen in Table 6.8, the LOI estimates are reasonable. For all five pathways, the LOI from the Ill. #6 is just over half that from Pit. #8. LOI from a staged wall-fired furnace is almost double the value from an unstaged furnace. And LOI-values for particles moving along gas ribbons and the quench layer are much lower than for the path through the furnace core in a T-fired furnace because the O2 levels are much higher throughout. These estimates suggest

Table 6.8 Predicted LOI for five furnace pathways. Path

Ill. #6

Pit. #8

Staged Wall Unstaged Wall T-Core T-Ribbons T-Quench

4.7 2.2 5.3 0.15 0.11

8.7 4.0 9.2 0.22 0.21

ChemNet furnace applications

249

that the assigned conditions for the five pathways give combustion histories that are at least qualitatively consistent with field test data. Distinct pathways through the furnace are within the scope of this analysis, not the full flowfield, so reactor networks that relax to plug flow are suitable. However, two aspects of this reaction system require special treatment. First, PAH spontaneously repolymerize into soot at flame temperatures which, given sufficient time, completely eliminates all PAH. Second, the O2 histories along all five paths contain sufficient O2 in some stages to eliminate all VOCs and PAH in tens of milliseconds at such hot temperatures. Production of both PAH and soot is described by the FLASHCHAIN®-based mechanism for tar decomposition (described in Chapter 7 of the first volume of this series). But neither the mechanism nor any test data characterize the very low levels of residual PAH for the sooting conditions in pc flames. This deficiency was bridged by retaining a portion of the predicted soot yield as the initial concentrations for three PAH species included in the homogeneous reaction mechanism: acenaphthalene, fluoranthene, and pyrene. Soot yields rather than secondary tar yields were used in the estimate because these two products differ only in their H-contents, and these differences are negligible on a mass basis. The MWDs for both primary tar and PAH extend to values well in excess of 1000 g/mole and have the extended tails toward large MWs seen in gamma distribution functions. Considering that the heaviest of the three representative PAH, pyrene, weighs only 202 g/mole, only a small fraction of the soot yield should be included as PAH in the initial volatiles mixture. This fraction was standardized to 0.10 for all calculations. Hence, the initial mole fractions for the three primary PAH species were specified from the relative yields of PAH and CH4, according to 0 yPAHi ¼ @X

1

  A xYPAH yCH4 YCH4 MPAH

MCH4 3 i¼1

(6.2)

i

where yPAHi is the mole fraction for each of the three representative PAH species; YPAH is the yield of PAH precursors, evaluated as the soot yield; Mi is the species molecular weight; Yi is the species yield in the secondary pyrolysis products in daf wt%; and x is the fraction of PAH assumed to remain after soot production, assigned as 0.10. According to Eq. (6.2), the mole fractions of PAH range from 10% to 25% of yCH4 for the three coals in this study. The near-instantaneous destruction of VOCs and PAH in the presence of O2 at flame temperatures is managed with two scenarios. In one, O2 was entirely eliminated from the process stream to give the maximum levels of VOCs and PAH at the furnace exit. This process stream comprised all the N2 in the injected air streams that met the overall furnace stoichiometry but none of the O2; plus all secondary volatiles including the three representative PAH species in the initial mixture; plus the oxidation products (CO, H2O, NO, SO2) of char and soot. The heterogeneous oxidation products were evaluated from separate simulations with the complete O2 histories in Fig. 6.36 for each path but were injected into a process stream that did not contain any O2. This idealization enables reforming chemistry involving CO, CO2, H2O,

250

Process Chemistry of Coal Utilization

and H2 that transforms VOCs and PAH without burning them away. In the second scenario, the same contributions to the process stream are supplemented by entrainment of either 10% or 20% of the O2 in the air streams, based on the entrainment histories behind the operating conditions in Fig. 6.36. In the ChemNet simulations, the first CSTR is fed by a mixture of coal moisture, gaseous secondary pyrolysis products, char, and the N2 in the air streams into the furnace. As N2 does not affect the VOC production chemistry, all of it is added into the first CSTR. The mean transit time through this CSTR is uniform at 250 ms to cover the complete devolatilization of the three coals. For each path, the total residence time beyond the 250 ms devolatilization zone was subdivided into unequal time increments that gave an extent of char conversion of 9% for each CSTR in the series, which gave about 10 CSTRs for the portion of the series in which char is converted. An additional six CSTRs were added to the series to resolve chemistry along the furnace convective passes into the economizer. According to the ChemNet simulations, there are substantial variations in VOC levels due to variations in both furnace pathway and coal quality. The persistent VOCs were ethylene, acetylene, ketene, benzene, indene, fluorene, and PAH. The qualified PAH were naphthalene plus substituted naphthalenes and naphthol; acenaphthalene; phenanthrene; anthracene; acephenanthrylene; fluoranthene; and pyrene. As the temperatures in the furnace simulations are very hot, there were also appreciable levels of many unstable intermediates in the predictions, particularly large PAH precursors. Most of these radical species would contribute to the VOC and PAH classes after quenching so they were included in the PAH yields. The VOC and PAH levels for the quench pathway without O2 entrainment for all three coals in Fig. 6.38 exhibit the expected dependence on coal quality, where PRB

VOC & PAH Levels, ppm

2500

PRB Pit8

2000 1500

VOCs III6

1000

400 300 200 100 0 0.0

Pit8 PRB III6

PAH Quench Stream 0.5

1.0

1.5

2.0

2.5

3.0

3.5

Time, s

Fig. 6.38 VOC and PAH levels from three coals for the quench pathway.

ChemNet furnace applications

251

denotes the subbituminous. For all coals, the VOC levels exceed 1000 ppm at the furnace exit and the levels with subbituminous and Pit. #8 are more-than-double that. For all three coals, levels of VOCs increase during the initial 1 s at the expense of diminishing PAH levels. This shift reflects the decomposition of large PAH into much smaller VOCs. Beyond 1 s, the levels of both VOCs and PAH are stable because the temperatures in this pathway become too cool for PAH destruction. Throughout a quench layer, the PAH levels are rank-ordered by the yields of GHCs plus PAH (as soot) from secondary volatiles pyrolysis (cf. Table 6.7): These sums are 25.8 daf wt% for Ill. #6; 27.9% for subbituminous; and 31.2% for Pit. #8. This correlation is not surprising because most of the PAH originates from these volatiles. In contrast, the VOC levels are greatest with subbituminous, and both subbituminous and Pit. #8 generate many more VOCs than the Ill. #6 at the furnace exit. The levels for the quench and ribbon pathways with Pit. #8 are compared in Fig. 6.39. Differences in temperature histories are crucial because all PAH are rapidly eliminated once the temperature exceeds about 1400°C. In the ribbon pathway, the temperature gradually approaches 1600°C during the first 750 ms, then maintains this level for another 1 s before it is slowly quenched (cf. Fig. 6.36). PAH elimination supplements the VOC inventory through 1 s. But once the PAH are destroyed, the VOC level abruptly diminishes and then continues to diminish at a slower rate throughout the rest of the simulation. This feature is a consequence of water gas shifting under highly reducing atmospheres. The only VOCs left once PAH had been destroyed were short-chain GHCs. Under reducing atmospheres, these hydrocarbons are stable for temperatures below 1200–1300°C but are oxidized at hotter temperatures as CO2 + H2 shifts into CO + H2O. The elimination of GHCs directly supplements the CO levels. In contrast, early into the quench layer the gas temperature approaches but does not exceed 1400°C then decays continuously. Consequently, only a small portion of PAH are converted into VOCs before both emissions saturate to relatively high levels.

2500

Quench

VOC & PAH Levels, ppm

1500

Quench

VOCs

2000

2000

VOCs

1500

1000

1000 Ribbons

Unstaged Wall

400

400

200

300

PAH

100 0 0.0

PAH

Quench

300

hv bituminous

Ribbons

Unstaged Wall

Quench

200

subbituminous

100

VOC & PAH Levels, ppm

2500

0 0.5

1.0

1.5

2.0 Time, s

2.5

3.0

3.5

0.0

0.5

1.0

1.5

2.0 Time, s

2.5

3.0

3.5

Fig. 6.39 VOC and PAH levels (left) from Pit. #8 hv bituminous for the (solid curves) ribbon and (dashed curves) quench pathways; and (right) from subbituminous for the (solid curves) unstaged wall and (dashed curves) quench pathways.

252

Process Chemistry of Coal Utilization

The same behavior is seen with subbituminous in the comparison between the unstaged wall and quench pathways in Fig. 6.39. Here, the transition to lower VOC levels is even more abrupt because the maximum temperature in the unstaged wall pathway exceeds 1700°C. This temperature is reached just after 500 ms, at which point the predicted levels of both PAH and VOCs plummet to zero. Compared to the case for Pit. #8, this case demonstrates that the elimination of PAH and VOCs accelerates for progressively hotter temperatures. Since GHCs are the primary remaining VOCs once PAH are destroyed and since the GHCs decompose into CO, the temperature dependences in the concentrations of VOCs and CO must be opposite: CO yields increase while VOCs diminish for progressively hotter temperatures above about 1300°C. This finding is important because it implies that PAH will not form in the core flows from any firing configuration, whether staged or unstaged, because such pathways have extended periods at or above 1600°C. Consequently, PAH can only form in the quench layers along the wall of any furnace type, simply because the temperatures along every other pathway exceed the PAH destruction threshold. This implication is especially secure because no PAH survives above 1400°C even without O2 entrainment, and O2 entrainment can only further accelerate PAH destruction. The results in Figs. 6.38 and 6.39 pertain to a hypothetical eddy that is initially fed by the products of secondary volatiles pyrolysis and then picks up the oxidation products of char and soot while it moves along one of the furnace flowpaths. But these calculations omitted O2 entrainment. The next calculation series has O2 entrainment rates of 10% and 20% of the injections of char oxidation products, which gave SR-values well under unity. The entrainment histories with Pit. #8 and Ill. #6 were very similar but that with subbituminous was faster because this char burns out faster than both other coals. As seen in Fig. 6.40, O2 entrainment dramatically reduces PAH but does not eliminate all of it. It also diminishes VOC levels but on a longer time scale. With all three coals, the ultimate PAH levels are just under 100 ppm, suggesting that only a family of especially stable PAH survives the partial oxidation. With Pit. #8 and PRB, the VOC levels with 10% O2 entrainment are reduced to about 10% of the initial level, whereas a quarter of the VOCs from Ill. #6 survive this entrainment level. The 20% entrainments with Pit. #8 and subbituminous leave only about 100 ppm VOCs, nearly all of which is PAH. The ChemNet analysis establishes several important features of the production of VOCs and PAH in furnaces. PAH pyrosynthesis is negligible compared to the conversion of primary tar into PAH because the levels of tar precursors are much greater than the reactants for pyrosynthesis, and the conversion efficiency of GHCs into pyrosynthetic PAH is well below 1%. The only furnace pathway that can deliver PAH into the furnace exit is the quench layer along watertube walls, because all other pathways have extended transit times through temperatures that are well above the threshold for complete PAH destruction, regardless of O2 availability. This explains why measured VOC levels are insensitive to firing configuration, as all furnace types have quench layers along their waterwalls. The predicted PAH levels were rankordered by the sum of yields of GHCs plus PAH in the secondary pyrolysis products. Even though the initial mixtures for Ill. #6 and subbituminous had much greater levels

2500

VOC & PAH Levels, ppm

2000

No O2

VOCs

1500 1000 PRB Quench Stream 400 +10% XCHAR

300 200 100

+20% XCHAR PAH

0 2500

III. #6 Quench Stream

VOC & PAH Levels, ppm

2000 1500 1000

No O2 VOCs

400 +10% XCHAR

300 200 100

PAH

0 2500

VOC & PAH Levels, ppm

2000 1500

No O2 VOCs

1000 Pit. #8 Quench Stream

400

+10% XCHAR

300 200 100 0 0.0

+20% XCHAR PAH 0.5

1.0

1.5

2.0 Time, s

2.5

3.0

3.5

Fig. 6.40 VOC and PAH levels from (top) subbituminous, (middle) Ill. #6, and (bottom) Pit. #8 for the quench pathway with O2 entrainments of (dotted curves) 0%, (dashed curves) 10%, and (solid curves) 20% of the char burnout requirement.

254

Process Chemistry of Coal Utilization

of VOCs and PAH than for Pit. #8, the exit levels of both emissions for Pit. #8 were more-than-double those for subbituminous, whose PAH levels were about double those for Ill. #6. Oxygen entrainment lowered PAH levels into the furnace exit and changed the speciation but did not eliminate them. It also reduced furnace VOC levels on a longer time scale. At the furnace exit, the predominant light VOC species were fluorene and acetylene, whereas the major PAH were unconverted initial PAH as fluoranthene and acenaphthalene. The predicted PAH speciation exhibits the established tendency for lighter PAH species to outlast the species with the greatest number of condensed rings. Larger PAH decomposed into smaller PAH and C1/C2 fragments including CH2O, CH2CO, CH4, C2H2, and C2H4, along with appreciable amounts of C6H6. The notable exception to this tendency is the persistence of fluorene and fluoranthene. CO does not mimic the decomposition of PAH or the VOCs in a quench layer with O2 entrainment because it is generated while both light VOCs and PAH are consumed in oxidation chemistry. Whereas PAH levels diminished for progressively hotter temperatures above 1300°C, CO levels increase in hotter furnaces.

6.3.3.3 The chemical structure of a 530 MW T-fired furnace This section presents a complete ChemNet analysis for a commercial coal-fired utility furnace. When fired with subbituminous coal, the furnace rating is 530 MW. The T-fired configuration uses 27 fuel and air injectors in each corner of the furnace within an elevation range of 11 m. In the operator’s nomenclature, each corner is denoted by a lower case letter, a, b, c, or d. In a plan view, the two lower corners from left to right are denoted as “a” and “b”. The two upper corners from left to right are denoted as “c” and “d”. There are 6 levels of injectors, each denoted by an upper case letter A, B, C, D, E, or F in order from bottom to top. The fuel injectors in the top level are out-ofservice. Each fuel injector is coupled to air jets above and below which are called “fuel air”. Each of these 3-jet assemblies is separated by an auxiliary airport. In addition, there are two CCOFA ports above the fuel injection registers. The coal flowrates to every fuel injector are the same. However, the air flows into fuel air and auxiliary air injectors are different among the corners and levels. Consequently, the flames emanating from different injectors have very different extents of fuel consumption and emissions, and most of these differences persist into the superheater section. In fact, once strong gradients in temperature and O2 level are established, they relax across the furnace elevation but do not disappear before the furnace exit. This furnace was analyzed with ChemNet for a single firing condition for which measured NOX and LOI emissions were available. Whereas the NOX emissions were matched by the predictions from the analysis, the overarching goal is not to simply predict furnace emissions. Rather, the goal is to reveal the chemical structure of the entire furnace; i. e., to determine where each major fuel component is released and burned away and which flow paths make the greatest contributions to NOX and LOI emissions. Given such detailed and heretofore unseen information, the operator hoped to identify more effective implementation and control strategies to achieve emissions lower than any recorded to date in commercial coal-fired furnaces.

ChemNet furnace applications

255

This presentation covers, in turn, the formal postprocessing of full CFD results to specify an equivalent reactor network; the marked differences in the operating conditions among different regions of the flowfield; the fuel conversion and emissions performance of each major path; and the main contributions to NOX and LOI emissions at the furnace exit.

Bulk flow patterns and chemically distinct regions The furnace was simulated, first, with conventional CFD to develop an equivalent reactor network and then with ChemNet simulations to generate flame structures and emissions predictions. The CFD simulations were performed by SmartBurn LLC with the chemistry submodel described in Chapter 4; the CFD results included the specialized output described in Chapter 5. The bulk flow patterns exhibit several unanticipated features and are among the most valuable elements of the ChemNet analysis. They were recognized by tracking particles or fluid elements, both inertial and massless, to identify the flow structures emanating from each injector. Whereas similar patterns were observed at each elevation, the magnitude of deflections and penetrations varied significantly from level to level. Each primary jet (of fuel plus primary air) is injected at about 45° from the corner walls next to the jet. Soon after injection, this flow is deflected, first, into a turn along the next sidewall by the jet from the previous corner, then, toward a sidewall by a swirling core on the vertical axis of the furnace. The momentum of the core becomes progressively stronger at each elevation, because it rotates faster to accommodate the additional fluid. Conversely, the extent of penetration of the primary coal jets toward the center diminish. Radiation from the walls and the core brings the primary jets to the onset temperature for devolatilization a few hundred milliseconds after injection. Then, the primary fuel jets are partitioned into separate streams by their interactions with the upstream jet fluid and with the core fluid. Due to their high inertia, nearly all the char particles and a small portion of primary air remain on their trajectory from the injector and move into the core. Most of the primary air and volatiles are stripped away into a separate stream that moves along the wall. This stream slowly mixes with air from the fuel air streams above and below the primary fuel jet, and burns. Primary jets do not retain their original structures or burn as isolated flames in this furnace. At an injector outlet, fuel particles move with their primary air stream and both fuel air jets as a coherent flow structure. Then the particle trajectories become relatively independent of the gas flows, due mainly to the interactions between a primary jet and the jet from the adjacent upstream corner. Most of the gas in the structure is directed toward the walls, but this change in direction pulls only a negligible fraction of the particles away from the primary suspension. The primary suspension continues toward the center, adding a layer to the core. Progressively more of the fluid in the original coherent structure is pushed toward the walls at progressively higher elevations, due to the growing momentum of the core. Whereas the gases are segregated, nearly all the particles penetrate into the core and radial gradients in the local suspension loading tend to be small.

256

Process Chemistry of Coal Utilization

The auxiliary air jets are also deflected toward the walls. They mix into the deflected gases from the primary fuel jets more gradually than the fuel air streams. As they contain no pulverized fuel, their momentum is significantly weaker than primary jets’ and they tend to be forced further toward the walls by the central core. The auxiliary air streams tend not to remain associated with their primary jets. Instead, they form a blanket of air along the furnace walls that rotates in an upward spiral. A rotating core begins to form from the interactions among the primary jets from the four corners at the lowest elevation, and expands as fluid is added at each successive higher injection elevation. The CFD simulations show that the core is radially stratified by the layers of fluids from each successive corner and each successive elevation from the bottom upward. This layered structure persists into the furnace exit. The layered flame structure indicates that only adjacent jets can mix with each other, whereas fluid from different elevations hardly interacts. The term “fireball” is often used to describe the core fluid in T-fired boilers. This term implies high swirl and fast rotation speeds which, in actuality, is misleading. Rotation speeds of the fireball fluid do increase at progressively higher elevations across the lower furnace. But the core fluid completes only one-half to one-andone-half rotations as it flows from the lower furnace into the superheat zone, depending on the elevations of its injectors. The core is only slightly enlarged by the first CCOFA injection. There is very little penetration into the core, so the bulk of the CCOFA moves along the walls with the air blanket from the auxiliary air injections at the lower elevations. Above the CCOFA jets, the swirl weakens along the upper furnace, and the fluid motion relaxes toward plug flow. There is very little gross mixing even through the convective passes. The weak mixing is important because the simulated concentration fields indicate very little O2 and an abundance of CO2 in the furnace core fluid. This reducing flow is surrounded by the vitiated air blanket and the products of volatiles combustion. Above the CCOFA elevation, O2 in the air blanket mixes into the core at a rate that largely determines the char burnout rate. The transverse O2 concentration profiles across the furnace at various elevations remain flat wherever char particles are present, which implies that the O2 transport rate is relatively fast compared to the aggregate burning rate. This situation is another consequence of the fairly low O2 level in the core throughout the upper furnace. The core flow in the upper furnace is surrounded by a quench layer that comprises the residual products of volatiles combustion and all residual air flows that did not enter the core. This layer contains no char particles, although it may have unburned soot from the primary fuel jets. The gas temperature diminishes sharply from the very hot core level to the much cooler gas temperatures along the furnace walls. The lateral profile of O2 concentration is similarly steep but inverted, because the highest O2 levels arise along the furnace walls. The temperatures of the furnace core at the lowest elevation vary from 1400 to 1550°C, and a small volume in the center of the core is hottest. At progressively higher elevations within the injection zone, the temperature of the core at each elevation varies from 1400 to 1600°C. Although the maximum temperatures at the center increase by only 50°C, the volume of the hottest gases increases due to the progressive

ChemNet furnace applications

257

penetration of the ignited fuel jets from the upper elevations. The gas temperature in the hottest region remains uniform. Above the CCOFA ports, gas temperatures in the core shells from each elevation vary from 1400 to 1620°C, but most of the furnace core is at 1620°C. A slightly cooler layer surrounds the hottest region. The temperature profiles toward the center change little as the core moves upwards into the superheater, but the size of the hottest region increases slightly due to the burning of char particles at the outer edge of the core and the entrainment of gases from the flow along the walls. The gas cools as the core moves into the superheat zone due to the convective heat transfer. The blanket flow surrounding the furnace core is several hundred degrees cooler, except for a thin layer adjacent to the core flow. The temperature difference persists well beyond the superheater. The field of combustibles mass fraction from Eq. (5.1) was analyzed with a threshold of 0.4 to delineate the fuel core for each fuel injector in the furnace, analogous to the fuel-rich core in the CRF pilot-scale flames (cf. Fig. 6.17). Fuel cores are generally smaller than their primary fuel jets because the combustibles are concentrated in the center of the jet, according to the normal distribution of species concentrations in cross-section due to entrainment of the surrounding fluid. A fuel core begins at some residence time after injection rather than at the injection plane. The intervening region between the injection plane and the beginning of the core is called the attachment region. Even though the combustibles mass fraction in the attachment region satisfies the threshold on the combustibles mass fraction, this region is marked as nonreactive on the basis of its cool particle temperatures. As explained above, the primary fuel jets separate into two streams once the jet flow is contacted by the flow from the upstream corner. Based on fluid and particle tracking, separation occurs within the fuel core of each jet after volatiles were released but before much of any of the fuels had burned away. “Ribbon” flows are distinguished by combustibles mass fractions greater than 0.4 (the value for fuel cores), due to the very high levels of volatiles, and are deflected toward the furnace wall. “Center” flows have combustibles mass fractions less than 0.4, even though these streams contain nearly all char and remain on the paths toward the furnace centerline. The fractions of primary air entrained in center flows vary from 11% to 39%. The higher entrainment fractions of primary air pertain to lower injector elevations because the furnace core better resists penetration as it grows into the higher elevations. Ribbons are regarded as mixing layers that constitute separate regions in the flowfield because they contain very high levels of gaseous fuels and soot. Each ribbon is surrounded above and below by two fuel air jets which, in turn, are surrounded above and below by two auxiliary air streams. These various air streams gradually mix with the ribbon flow, burning out the volatile fuels, and also with each other. As seen in Fig. 6.41, the ribbon from each primary jet moves in a helix from the fuel core of its parent fuel jet to the top of the fuel injection zone. The air flows between the ribbons from different fuel jets tend to isolate each ribbon, and interaction among the ribbons from primary jets at different corners and elevations is very weak. Neither the CCOFA injectors nor the turn into the convective passes disperses the ribbon flows into the furnace core of the flowfield.

258

Process Chemistry of Coal Utilization

Fig. 6.41 Visualization of the ribbon flows from five injector elevations through an imaging plane at the CCOFA elevation.

The furnace core in the center of the flowfield constitutes a single region that contains all char particles and small portions of the air injected with the primary fuel jets. Char burnout, either through char oxidation or, perhaps, char gasification by CO2 and H2O, is the dominant chemical process in the furnace core, because all the volatiles were swept into the ribbons. Residence times in the core of char injected at the lowest elevation are significantly longer than those for char injected at the highest elevation. The furnace core was not significantly disrupted by the two levels of CCOFA jets because they follow flow patterns like auxiliary air jets in the injector zone. There is little direct penetration of the CCOFA jets into the furnace core. Consequently, the structure of the furnace core persists through the CCOFA elevations into the convective passes. No large-scale mixing disperses this flow into the ribbon flows near walls. Instead, the core flow turns in near-solid body motion into the convective passes and proceeds into the furnace exit. It remains significantly hotter than its surrounding blanket flow throughout this trajectory. Above the CCOFA injectors, the furnace core is surrounded by an O2-rich, relatively cool, helical flow near the walls called the “quench layer.” It comprises volatiles combustion products, vitiated fuel air and auxiliary air, and CCOFA. It is the source of

ChemNet furnace applications

259

O2 that is entrained via turbulent diffusion into the furnace core to sustain char burnout. Oxygen is transported along with moisture and N2 whose concentrations are also significantly different from their levels in the core. While this region loses O2 to the furnace core, its chemistry adapts to the more reducing environment by eliminating NO and generating CO, as long as the temperature stays hot enough to sustain the oxyhydroxyl radical pool. Residual soot in the volatiles combustion products may also be burned away in the quench layer. This region cannot be specified from a threshold on combustibles mass fraction, because these values in the upper furnace are fairly uniform and low-valued. (Recall that the definition of combustibles mass fraction in Eqs. (5.1a-c) does not distinguish combustible elements in products from reactant species.) It is delineated by large gradients in the transverse O2 concentration profile across this region, as seen in Fig. 6.42. This figure shows the transverse profiles of gas temperature and O2 mass fraction at several elevations in the upper furnace, beginning with the CCOFA injection elevation and ending with the inlet to the superheater. Both gas temperature and O2 concentration are essentially uniform across the furnace core, then change in opposite directions across the quench layer to their respective values on the furnace walls. The O2 concentrations maintain the same transverse profiles while diminishing toward zero at the highest elevations. The maximum O2 levels on the furnace walls become substantially lower for progressively higher elevations. As there are hardly any fuel compounds in the quench layer, the features in Fig. 6.42 are interpreted as withdrawal of O2 from the quench layer into the furnace core via turbulent diffusion, so that any species whose concentrations in the core and quench layer are substantially different will be transported at the same rate. As the O2 concentration profiles do not shift with elevation, the O2 withdrawn from the quench layer into the core is consumed relatively slowly by char oxidation, which enables the profiles to equalize.

0.18 1800

0.16 0.14 O2 Mass Fraction

Gas Temperature, °C

1600 1400 1200 1000 Heights above SOFA, m 800

0 5 10 15

0.1 0.08 0.06 0.04

0 5 10 15

0.02

600 –10

0.12

Heights above SOFA, m

0 –5

0 Distance from Centerline, m

5

10

–10

–5

0

5

Distance from Centerline, m

Fig. 6.42 Transverse profiles of (left) gas temperature and (right) O2 mass fraction at four elevations in the upper furnace.

10

260

Process Chemistry of Coal Utilization

The extent of the furnace core is evaluated as the position where the gas temperature equals 1427°C because it corresponds with the position where the O2 level approaches zero on the right side of the furnace at all elevations in the upper furnace, and also with the outer boundary of the two-phase suspension flow. Based on this definition, the extent of the core hardly contracts for progressively higher elevations. The outer boundary of the quench layer is positioned on the furnace walls. The gas temperature falls by several hundred degrees from the outer edge of the core flow to the furnace walls. This is problematic because, by definition, each region can only be assigned a nominal average thermal history along the flow direction. Accordingly, the quench layer is subdivided into a high-temperature layer adjacent to the furnace core and a low-temperature layer adjacent to the furnace walls. The high-temperature layer feeds fluid into the low-temperature layer until it vanishes at the superheater. Along the way, O2 is entrained through the quench layer into the furnace core at rates evaluated from the CFD flowfield. To this point, the furnace containing twenty fuel jets has been subdivided into the following five regions: Each of these jets is initially subdivided into (1) an attachment region (based on a combustibles mass fraction over 0.4 and mean temperatures below 450°C) and (2) a fuel core (based on combustibles mass fractions of 0.4). Each primary jet then partitions into two flows. Most of the primary air plus volatiles form (3) a ribbon that spirals upward toward the CCOFA elevation (based on combustibles mass fractions over 0.4 and streamlines that diverge away from the char suspension). Ribbons are mixing layers, entraining fuel air and auxiliary air. The rest of the primary air plus char particles move toward the center into a (4) central furnace core that expands as it grows by additions at each of the higher injection elevations. There is little mixing among the contributions from each elevation, so the core is stratified into five shells that persist through the CCOFA elevations and moves in near-plug flow through the convective passes. Above the highest CCOFA elevation, the core is surrounded by (5) a quench layer (based on boundaries at gas temperatures of 1427°C and the furnace walls). The quench layer comprises all CCOFA air, vitiated air associated with the ribbons, and the products of volatiles combustion. It is the source of O2 transported into the core that sustains char oxidation along the furnace core. As it loses O2, its progressively more reducing chemical environment adjusts the NO and CO levels to their ultimate contributions to the furnace exit levels. It is subdivided into two subregions at approximately 1050 and 1450°C. A schematic for these assignments appears in Fig. 6.43 where each region appears as a rectangular box. The larger dot-dash box in the lower left represents the regions for a single primary fuel jet. There are 20 such assemblies in the entire furnace, although only one appears in Fig. 6.43, and they are not identical. Each assembly contains an attachment region and fuel core that are fed by the primary fuel and air streams. The attachment region is nonreactive whereas the fuel core sustains devolatilization but not combustion. A small amount of fuel air is entrained into the fuel core before its splits into a ribbon and center stream. The ribbon contains most of the volatiles and primary air. It also entrains most of the fuel air and auxiliary air as it moves toward the CCOFA injectors, burning out the volatiles. The center stream feeds char and a minor portion of the primary air into the central furnace core.

ChemNet furnace applications

261 Injection Zones A B C D E

Upper Convective Furnace Passes

Five Shells of the Central Core O2 Char + 1° Air

1° Air + Fuel

Attachment Length

Core

Exhaust Quench Layer

Volatiles +O2

Fuel Air + AUX Air

Ribbon

CCOFA

Fig. 6.43 Schematic for the assigned regions in the furnace.

The central furnace core forms at the lowest injector elevation and persists through the convective passes. This region contains all the char, some primary air, and small fractions of fuel air and auxiliary air, which are not depicted in Fig. 6.43. It sustains char oxidation and, perhaps, char gasification by CO2 and H2O in the lower furnace zone, and char burnout through the upper furnace and convective passes. As the primary air is quickly depleted, most of the char burns in the upper furnace at a rate largely determined by the entrainment rate of O2 from the quench layer into the core. The O2 comes from the aggregate flow of ribbons, residual fuel air, and auxiliary air, plus the CCOFA flows into the quench layer. The quench layer begins at the second CCOFA injection elevation. It comprises volatiles combustion products and vitiated air in the ribbons plus residual fuel air and auxiliary air plus all CCOFA. While the quench layer loses O2 (and N2 and H2O) to the core, its composition adjusts to the progressively more reducing environment. At some point downstream of the superheater, the quench layer mixes with fluid in the furnace core to form the whole flue gas. However, the CFD simulation indicated that the quench flow remains on the walls through the convective passes, even though the corner above the upper furnace contains fairly large recirculation zones on two sides. Char continues to burn only in the furnace core along the convective passes. Fuel air, auxiliary air, and CCOFA are fed into regions by entrainment, but do not constitute regions because intermixing of these various air streams does not sustain any chemistry. No fuel compounds are present while the fuel air and auxiliary air streams mix. Similarly, the volatiles have already burned in the ribbons by the time they reach the CCOFA injectors.

Equivalent reactor network The mean residence times and CSTR-numbers assigned from the CFD-based RTDs for all near-injector regions are collected in Table 6.9. Only the residence times are reported for the attachment lengths because this region does not sustain chemistry, so RTDs were not required. This time lag varies from 118 to 189 ms, but not in

262

Process Chemistry of Coal Utilization

Table 6.9 Residence times and CSTR-numbers for near-injector regions. Attach.

Fuel Cores

Ribbons

Elev.

Corner

τ, s

τ, s

NCSTR

%1°Air

Split

τ, s

NCSTR

A

a

0.135

0.212

8

62.1

b

0.121

0.310

23

63.0

c

0.138

0.210

9

68.1

d

0.147

0.198

11

60.6

a

0.137

0.208

27

68.5

b c

0.150 0.139

0.185 0.217

14 4

88.3 74.1

d a

0.139 0.159

0.206 0.270

15 4

81.6 87.8

b

0.150

0.186

35

70.2

c

0.132

0.208

8

83.4

d

0.181

0.184

16

78.6

a

0.165

0.236

12

85.2

b c

0.159 0.118

0.215 0.180

22 3

85.1 81.5

d

0.126

0.146

11

87.4

a

0.139

0.234

8

89.1

b

0.176

0.198

15

86.1

c d

0.140 0.189

0.206 0.246

3 4

78.7 66.1

45.6 29.6 24.8 51.6 48.4 64.5 35.5 19.6 80.4 74.7 25.3 100 22.2 77.8 100 28.9 71.1 40.5 59.5 19.9 80.1 50.5 49.5 38.3 61.7 100 18.1 81.9 36.5 32.2 31.3 50.4 49.6 46.2 53.8 100 100

0.867 1.020 1.120 1.080 1.260 0.869 0.959 0.757 1.290 0.534 0.682 0.907 0.092 0.695 0.794 0.015 0.413 0.695 0.879 0.104 0.476 0.456 0.593 0.014 0.296 0.582 0.018 0.252 0.282 0.355 0.466 0.025 0.160 0.093 0.190 0.096 0.408

124 126 126 126 126 126 126 98 126 69 126 126 1 52 43 1 9 126 126 2 20 126 118 2 8 27 2 123 124 126 126 8 8 24 42 5 4

B

C

D

E

ChemNet furnace applications

263

any systematic way with the elevation of the injector. Similarly, the mean residence times for the fuel cores are highly variable, from 180 to 310 ms, due to the different shapes of the various fuel cores which, in turn, were determined by the interactions between the primary jets from adjacent corners at the same elevation. These interactions were neither symmetric in the furnace cross-section nor uniform from injector to injector. Consequently, the numbers of CSTRs assigned for cores vary from 3 to 35, which spans highly disperse RTDs through plug flow. For the reaction mechanisms in the flame simulations, the impact of different CSTR-numbers is not noticeable for series that have 20 or more reactors. Nonetheless, different extents of combustion and, especially, different extents of fuel-N conversion can be anticipated among the fuel cores. The results for ribbons show the percentages of primary air swept from the primary fuel jets into ribbons, as well as the mean residence times and CSTR-numbers. As expected, there is less penetration of primary air into the furnace core at higher elevations. Whatever portion does not penetrate remains with its ribbon flow. Ribbon RTDs were usually too complicated to represent with a single CSTR-series. All but five of these flows have RTDs with more than one mode. Thirteen were bimodal and two were trimodal. The proportions of the total flow in each mode are reported as the split percentages in Table 6.9. These RTDs were evaluated with particle tracking as the transit time between the origin of the ribbon at a core and a plane over the entire furnace cross-section just downstream of the CCOFA jets. The fluid that started in a ribbon may have passed through the CCOFA elevation far away from the bulk of the fluid that started with that same ribbon. The RTDs of the five shells of the furnace core indicated plug flow so this region is represented by five CSTR-series of at least 30 reactors each. The mean residence times from the CFD results were 3.09, 2.84, 2.48, 2.28, and 2.28 s, respectively, for progressively higher elevations. Two RTDs were assigned for the quench layer, one for a high-temperature inner layer that vanishes at the superheater and a second for a cooler layer on the furnace walls. The high-temperature layer moves in plug flow with a mean residence time of 1.06 s. As fluid enters the low-temperature layer from its inside boundary, the RTD has much greater dispersion. This flow is well-represented with six CSTRs and a mean residence time of 2.84 s. The mean gas temperatures at the end of attachment lengths varied from 375°C to 450°C due to the coarse resolution in the CFD temperature field in these regions. The mean gas temperature histories for the fuel cores on the lowest and highest fuel injection elevation appear in Fig. 6.44. Generally, they are within a 100°C band except for the histories for injectors dE and bA, which are significantly cooler after 100 ms. Wall temperature histories for the fuel cores were uniform at each elevation, because they were assigned as the average gas temperature over the furnace cross-section at each fuel injection elevation. They varied from 1120°C to 1180°C. The mean particle temperatures for the devolatilization simulations for the fuel cores at elevations A and E also appear in Fig. 6.44. The heating rates are fairly uniform and never exceed 104°C/ s, whereas ultimate temperatures vary from 1000°C to 1250°C. Coal suspensions at commercial loadings do not heat nearly as fast as isolated coal particles in the p. f. size grade.

264

Process Chemistry of Coal Utilization 1400

1200

Mean Gas Temperature, °C

1000

1000

dE

dE

800 800 600 600

400 Flame Cores Elev A: Dashed Elev E: Solid

400 0.00

0.05

0.10

0.15

0.20

Time, s

0.25

0.30

0.35 0.00

Flame Cores Elev A: Dashed Elev E: Solid 0.05

0.10

0.15

0.20

0.25

0.30

0.35

Mean Particle Temperature, °C

1200

bE bA

200 0

Time, s

Fig. 6.44 Temperature histories for (left) gas and (right) particles for the fuel cores at elevations (dashed curves) A and (solid curves) E.

Surprisingly, the entrainment of fuel and auxiliary air streams into fuel jets was remarkably weak. The eight fuel cores from elevations A and E entrained less than 4% of air with only one exception, which entrained less than 10%. This is a direct consequence of the extremely heavy coal loadings in commercial furnaces, which fall between 0.4 and 0.5 kg-coal/kg-primary air. Such momentum-dominated jets are extremely difficult to disperse, as evidenced in this furnace by weak secondary air entrainment and deep penetration of char particles into the furnace core. Clearly, the fuel cores burn under very rich stoichiometric ratios, which will inhibit NOX production provided that residence times are sufficient to burn out the volatile fuel species. Indeed, in a recent ChemNet analysis of only these fuel cores, no NO at all was present in the effluent from fuel cores for the commercial coal loading (Niksa, 2019c). Among all regions, conditions are most variable for ribbons, as seen in Fig. 6.45 for the lowest and highest fuel injection elevations. Maxima in the gas temperature histories vary from 1450 to 1600°C, and at least 300 ms elapses before the maximum gas temperature is achieved with ribbons from elevation A. This period is sufficient for the conversion of most of the volatile-N species. However, the ribbons from top elevation E only last for 80 to 175 ms, and will carry HCN and NH3 into the upper furnace, where they are converted at higher O2 levels and cooler temperatures in the quench layer. The O2 entrainment histories for ribbons in Fig. 6.45 closely correspond with the gas temperature histories. Two ribbons from elevation A entrain all the available fuel air and auxiliary air in roughly 225 ms and have the hottest temperature histories because the mixing rates of secondary air with the volatiles in the ribbons determines the volatiles burning rates in the CFD simulations. Two others have stepwise entrainment histories, which pushes the gas temperature histories through a cooler maximum temperature after the first interruption in the entrainment history. Beyond that point, additional entrainment only diminishes the cooling rates in the gas temperature histories. The ultimate entrainment levels are similar for both groups

ChemNet furnace applications

265 1.0

0.8

1500 1400

0.6

1300 0.4 1200

1000 0.00

Ribbons Elev A: Dashed Elev E: Solid

Ribbons Elev A: Dashed Elev E: Solid

1100

0.25

0.50

0.75

1.00

1.25

1.50 0.00

0.25

0.50

Time, s

0.75

1.00

1.25

Entrainment, %

Mean Gas Temperature, °C

1600

0.2

0.0 1.50

Time, s

Fig. 6.45 Mean histories for (left) gas temperature and (right) air entrainment for the ribbons from elevations (dashed curves) A and (solid curves) E.

because all the available secondary air flows are ultimately entrained into ribbons. Similarly, at elevation E the ribbons with the fastest entrainment rates burn to the hottest gas temperatures, as expected. Mean gas temperature histories for the furnace core shells from the five injection elevations appear in Fig. 6.46. The histories are plotted from their time of injection, so they appear to end at different times but are actually staggered initially but coincide at the reheat zone outlet. Each history increases rapidly from the initial value at the split point in its parent fuel core, which is about 750°C. They then increase to fairly uniform values in 300 ms. The maximum gas temperature diminishes for progressively higher elevations but by no more than 150°C among all shells. All histories are rapidly cooled by the superheater and quickly fall below the threshold for ignition during the latest stages of burnout. Wall temperature histories for the core shells were assigned as the average temperatures of the adjacent shells. They are similar in form to the gas temperature histories and even more uniform. Entrainment histories for furnace core 1600 1500 High-T Layer

1400

1400 1300

1200

1200

Low-T Layer

1100

1000 EDC

800

600

0.0

B A

1000 900

Furnace Core Shells 0.5

1.0

1.5

2.0

Time, s

2.5

3.0

Quench Layer 3.5 0.0

0.5

800 1.0

1.5

2.0

2.5

3.0

Time, s

Fig. 6.46 Mean gas temperature histories for (left) furnace core shells from elevations A through E and (right) two streams in the quench layer.

Mean Gas Temperature, °C

Mean Gas Temperature, °C

1600

266

Process Chemistry of Coal Utilization

shells increase in proportion to transit time and saturate to approximately 20% of the fuel air, auxiliary air, and CCOFA streams. The mean gas temperature histories for the two streams in the quench layer appear in Fig. 6.46. The high-temperature layer persists for just over 1 s, and remains hotter than 1450°C throughout. The low-temperature layer never gets hotter than 1250°C and takes almost 3 s to pass into the reheat zone. The entrainment history imposed on the furnace core above the CCOFA elevation was rescaled to the slightly different time scale in the quench layer, and then implemented as the turbulent diffusion rate from the quench layer. To achieve the very high extent of burnout expected for the subbituminous coal, the entrainment rate from the CFD simulation had to be increased by a factor of 1.6. In summary, the operating conditions in the regions near injectors are remarkable for their diversity. Such variability is rooted in the interactions among the fuel jets from a single elevation, and among the jets from different elevations. It is an essential feature of the performance of this furnace. Ribbons sweep away 60% to 90% of the primary airstreams; have residence times that diminish for progressively higher elevations from well over 1 s to under 200 ms; and maximum gas temperatures ranging from 1400 to 1600°C. All ribbons entrained over 80% of the available secondary air, but the entrainment rates were highly variable. The entrainment rates determined the form and maximum values in the gas temperature histories, consistent with the mixing-limited rates of volatiles combustion in the CFD simulation. Each of these variations would significantly affect the conversion of volatile-N species into NO. But the fact that almost all the gas temperatures in ribbons increase through moderate values during the initial 250 to 300 ms of residence time is a potentially important mitigating factor on NOX production. Conditions in the shells within the furnace core are much less variable. Residence times in the core (into the reheat zone) are under 3 s, except for the fluid trapped in the recirculation cells at the inlet to the convective passes. Maximum gas and wall temperatures vary by only 150°C. Only the entrainment of primary air into core shells was significantly variable, ranging from 5% to 15%. Whereas the flow in the quench layer was very uniform, it sustains steep gradients in both O2 concentration and gas temperature. The temperature difference was coarsely resolved in terms of two streams, one whose mean gas temperature was at least 300°C hotter than the cooler layer on the furnace walls.

Near-injector flame structures With the reference subbituminous coal, the main volatile fuel components are soot, CO, H2, and GHCs as CH4 and C2H2. Ultimate volatiles yields are approximately 70 daf wt%. The only volatile-N species is HCN, which accounts for almost 80% of the coal-N. The remainder is in the char. The predicted flame structures of a fuel core and gas ribbon at injector Bb appears in Fig. 6.47. In descending order, the three panels display the variations in the gas temperature and SR-values for the gas phase only; the mass fractions of O2 and CO; and the mass concentrations of the major N-species. Each parameter is plotted versus the

ChemNet furnace applications

267 20

10

SR

800 700 600

5

TGAS

1500

1.25 1.00

1400

0.75

1300

Entrainment Fraction

0.50 TGAS

1200

0.25 0.00 3.0

0

400 CO

2.0

15 Injector Bb

10

100 XSOOT

2.5

O2

1.5 1.0

5

CO, % by mass O2 & CO, % by mass

20

O2

CO

2.5

80

2.0 1.5

Gas Temperature, °C

15

60 Ribbon Bb 40

Soot Burnout, %

SR

900

500

O2, % by mass

1600

SR

Ribbon Bb 1.50

1000 Gas Temperature, °C

1.75

Injector Bb SR & Entrainment Fraction

1100

1.0

0.5

0.5

0.0

0.0

20

0 0

Injector Bb

Ribbon Bb 1600

150

1600

HCN

1200

1200 100 800 50 400

0 0.00

NH3 0.05

0.10 Residence Time, s

0.15

0 0.20

100

NO

800

50 400

NH3 0 0.0

HCN & NH3, ppmw

NO, ppmw

HCN

HCN & NH3, ppmw NO, ppmw

NO 150

0 0.2

0.4

0.6

0.8

1.0

1.2

Residence Time, s

Fig. 6.47 Structure of (left) core Bb and (right) ribbon Bb showing (top) the operating conditions, (middle) major species, and (bottom) N-species.

mean residence time. Since all these characteristics pertain to the gas flows, the adjacent panels should be read sequentially, where this ribbon is a continuation of this fuel core. However, some quantities appear on different scales in adjacent panels. For this particular injector, devolatilization is completed within 170 ms, the flow leaves the fuel core at 183 ms, and the ribbon flow moves to the CCOFA elevation at 1.09 s. Neither H2 nor GHCs are present in fuel cores in significant amounts. The H2 mass fraction stays within 500 to 1500 ppmw throughout the first 600 ms then vanishes from the ribbon. The only interval when GHCs are present is from 50 to 70 ms, when the C2H2 concentration reaches 3000 ppmw. GHCs ignite the flow but are otherwise unimportant. They are certainly not effective NOX reductants. The gas temperature steadily increases along the fuel core. If the coal-based SRvalue of only 0.35 was operative, N-species would be converted under extremely rich conditions in fuel cores. However, the SR-value for the gas phase begins at infinity, the nominal value for pure primary air, and then falls sharply while volatiles are

268

Process Chemistry of Coal Utilization

released into the flow, making it more reducing. But it does not even approach the whole-coal value despite the abundant yield of volatiles from this coal, because a very large portion of volatiles are converted into soot, which does not factor into the SRvalue for the gas phase. Even at the end of devolatilization, the SR-value is 0.95, which is almost three times larger than the whole-coal-based value. The chemical environment in the fuel core is decidedly oxidizing. The volatiles ignite at roughly 650°C, as evident in the decay in the O2 concentration and the surge in the CO concentration. At this point, only 10% of the ultimate volatiles yield has been released. All accumulated GHCs are consumed at ignition, and these concentrations remain very low throughout. Due to the relatively low temperatures across fuel cores, soot passes into the ribbon unignited. The NO concentration initially surges to 170 ppm due to the rapid conversion of HCN, the primary volatile-N species, in the lean section of the core where the SRvalue falls from 5 to unity. But once O2 falls below 5%, the NO concentration diminishes. The NO reduction stage coincides with the appearance of NH3 and with a surge in the HCN concentration. The effluent from the fuel core contains 1600 ppm HCN and 250 ppm NH3 but only 75 ppm NO. Within this chemical environment, the identity of the primary volatile-N species is inconsequential, because homogeneous chemistry rapidly interconverts any volatile N-species into a mixture of HCN, NH3, and NO in proportions determined by the local SR. In addition to the N-conversion chemistry, the NOX reduction zone sustains sufficient water-gas shifting to nearly double the CO concentration. Profiles through the ribbon from injector Bb also appear in Fig. 6.47. This region has the hottest temperatures in the flame, plus nearly all the fuel air and auxiliary air streams. The gas temperature continues its steady increase to an intermediate plateau at 1540°C then rises more gradually to 1600°C at the ribbon outlet. The SR-value for the gas phase remains reducing until 580 ms then surges above 1.6 with the addition of fuel air and auxiliary air from 580 to 860 ms. The close parallel between the SR-value and the air addition rate is clearly seen in the cumulative air entrainment fraction in Fig. 6.45. As soon as secondary air is entrained, the SR-value in the gas phase exceeds unity and remains firmly in the oxidizing regime for the rest of this flame. The CO burns slowly until the entrainment rate surges at 580 ms. It is completely consumed by 900 ms. Similarly, the soot ignites at the ribbon inlet but does not burn at a rapid rate until 580 ms, when the gas temperature exceeds 1500°C. Thereafter, all combustion is complete so the O2 concentration surges to 2.8% and remains at this level throughout the rest of the ribbon. Fortunately, residual HCN and NH3 from the fuel core are eliminated while the gas phase is reducing so only 5.4% of these volatile-N species are converted into NO. Before the surge in the air entrainment rate at 580 ms, the NO concentration falls to only 10 ppm while the NH3 level grows to 600 ppm, and the HCN level falls below 1000 ppm. Note that almost all the eliminated HCN and NO are converted into N2 during this stage. But the surge in air entrainment shuts down NO reduction before the fixed nitrogen species can be fully converted into N2. Both NH3 and HCN are rapidly burned away before 700 ms, and the NO concentration grows to 100 ppm in the ribbon effluent. The delayed entrainment of fuel air and auxiliary air at this injector is responsible for extensive NO reduction and the relatively low ultimate NO concentration at the end of the ribbon.

ChemNet furnace applications

A

B

269

C

D

E

Ribbon NOx, ppmw

Ribbon O2, mass %

B

C

D

E

A

B

C

D

E

250

4

3

2

1

200

150

100

50

0 30000

A

B

C

D

0 800

E

25000

Ribbon HCN and NH3, ppmw

20000 15000

Ribbon CO, ppmw

A

10000 5000 1000

500

0

D

HCN NH3 600

400

200

0

E

0

5

10

15

20

25

Injector Index

Soot Burnout, %

80

60

40

20

0 0

5

10

15

20

25

Injector Index

Fig. 6.48 Ribbon effluent levels of (left) (top) O2, (middle) CO, and (bottom) soot burnout; and (right) (top) NO and (bottom) HCN and NH3. Injector elevations from bottom to top are denoted by upper case letters A–E, and for each elevation, the set of four bars are for corners a, b, c, and d from left to right.

The structures of the other near-injector regions are developed from the same transitions, except that variations among the entrainment rates, temperature histories, and RTDs significantly affect the emissions. The variations in the combustion characteristics are apparent in Fig. 6.48, which shows the concentrations of O2 and CO, extents of soot burnout, and the N-speciation in the ribbon effluents from all twenty injectors. The overall tendency is for nearly complete volatiles combustion in the ribbons from

270

Process Chemistry of Coal Utilization

elevation A to mostly incomplete combustion in the ribbons from elevation E. In the ribbons from elevation A, the residual O2 levels are all greater than 2%; the CO levels are never more than a few hundred ppm; and the soot is completely burned away. In contrast, the ribbons from elevation E never contain as much as 1% residual O2, but their CO levels range from 2500 to almost 30,000 ppm, and the extents of soot burnout range from 18% to 71%. The main reason for these variations is the much shorter residence times available in the ribbons from the higher elevations, which are insufficient to mix most of the fuel air and auxiliary air into these ribbons. The performance could be improved by accelerating the entrainment rates in the upper injectors, especially regarding CO concentrations. Whereas volatiles combustion is essentially complete in all ribbons from elevation A, all the other elevations include cases with incomplete combustion. Such cases are characterized by relatively low residual O2, high CO emissions, and low extents of soot burnout. This pattern is apparent in the emissions from ribbons Ba, Bc, Ca, Cc, Da, Dc, Ea, Eb, Ec, and Ed. Ribbon Bc is especially interesting because its poor volatiles combustion efficiency is due to a short circuit that moves nearly a quarter of the flow to the CCOFA elevation in only 92 ms, vs. almost 700 ms for the remainder of the flow. Recall that all the injectors were in the injection plane at the same horizontal firing angle with the same coal flowrate in the CFD simulation. Clearly, uniform injector settings do not ensure the most uniform combustion behavior, due to the interactions among the flows from adjacent injectors. It remains to be determined whether or not injector fine-tuning can enhance the ribbon combustion efficiency, although this analysis predicts a large potential payoff albeit within the uncertainties on the CFD simulation. The impact of the intrinsic variability on the N-speciation also appears in Fig. 6.48. The tendency is for progressively higher levels of the fixed-N species (HCN and NH3) for progressively higher elevations, and somewhat lower outlet NO levels. However, it would be incorrect to expect lower ultimate NO levels for the ribbons from the higher elevations because the fixed-N species will definitely be converted into both NO or N2 along the quench layer. In fact, this competition will be shifted toward significantly greater NO emissions for the ribbons from higher elevations because it will take place where O2 concentrations are elevated by the injection of CCOFA and the entrainment of all residual fuel air and auxiliary air. As a means to reduce NO emissions, the residence times in the ribbons from the upper elevations should be extended to enable the fixed-N species to be reduced more completely. Accelerating the entrainment rates of fuel air and auxiliary air would not be effective because, as seen in the structure of ribbon Bb (in Fig. 6.47), air entrainment must be delayed until the fixed-N species have been reduced to minimize the ultimate NO emissions. The five shells of the central furnace core sustain char oxidation but no volatiles combustion or soot chemistry. Nevertheless, chemistry in the gas phase profoundly affects the ultimate emissions from this region. The shells are characterized in two parts, a lower furnace zone which remains very reducing after all the O2 in primary air has been rapidly consumed, and an upper furnace zone which entrains much larger portions of O2 from the quench layer. The CCOFA elevation separates these parts.

ChemNet furnace applications

1700

0.8

1400

TGAS

0.6

Entrainment Fraction

1200

0.4 1000 0.2

Core Shell A To CCOFA

0.0

800

4000 3000

Core Shell A To CCOFA

1100 Entrainment Fraction Core Shell A CCOFA to Economizer

1000 900 O2

15000 10000 CO 5000

50

500

40

400

Core Shell A CCOFA to Economizer

XCHAR

400 Core Shell A To CCOFA

300

30 20

200

NO

NO, ppmw

NO XCHAR

Char Burnout, %

NO, ppmw

1200

0

500

90 80

300

70

200

60

100

50

10

100 0 0.0

1300

Char Burnout, %

0 600

1400

Core Shell A CCOFA to Economizer

0.2

2000 1000

TGAS

0.4

20000

CO

O2 or CO, ppmw

O2 or CO, ppmw

O2

0.6

1600 1500

0.8

0.0

6000 5000

SR

1.0

Gas Temperature, °C

1600

SR

Gas Temperature, °C SR & Entrinment Fraction

SR & Entrinment Fraction

1.0

271

0.2

0.4 0.6 0.8 Residence Time, s

1.0

0

0

1.0

1.5 2.0 2.5 Residence Time, s

3.0

Fig. 6.49 (Top) Conditions, (middle) O2 and CO, and (bottom) NO and char burnout in furnace core-shell A across the (left) lower and (right) upper furnace, where the scales for residence time are different.

The chemical structures of the innermost shell over both levels of the furnace core appear in Fig. 6.49. Here too the adjacent panels should be read sequentially, albeit on different scales in adjacent panels for some quantities. Across the lower furnace core shell, the gas temperature increases rapidly during the first 400 ms, when it reaches a plateau near 1600°C. As the char is initially entrained in primary air, the SR-value for the gas phase begins at infinity, like the values in the fuel cores. It then falls sharply while char oxidation products are released into the flow, making it more reducing. Thereafter, the SR-value remains very slightly richer than unity, even while small portions of the fuel air and auxiliary air are entrained into the flow. The reason that SR remains constant is that each addition of O2 is rapidly converted into CO2 by char oxidation and homogeneous CO oxidation, which perturbs the SR-value toward unity. Most importantly, the chemical environment in the furnace core is much more reducing than expected. The O2 in the furnace core falls from 21.1% by mass to less than 1000 ppmw in less than 100 ms. The O2 concentration remains below a few hundred ppm throughout the lower furnace. Despite the trace levels of O2, the extent of char

272

Process Chemistry of Coal Utilization

burnout increases by 15% after the O2 in the primary air has been consumed due to the entrainment of fuel air and auxiliary air. Each addition is very rapidly consumed in char oxidation. The NO concentration initially surges to 600 ppm while more than a third of the char is burned out by the O2 in primary air. This amount is determined by the conversion factor for char-N, which was set to 0.2 in the ChemNet simulations (so that 20% of the char-N is released as NO in proportion to the char burning rate). But after the initial surge in the NO concentration, chemistry in the gas phase quickly reduces away the NO. The NO reduction persists throughout the lower furnace. Consequently, the NO concentration increases from 17 to only 87 ppm while the extent of char oxidation increases by 15%. Based on the conversion factor for char-N, the increase would be nearly 300 ppm if gas-phase chemistry did not mitigate the NO level. It appears that the CO and H2 generated by water gas shifting at the high stream temperatures are responsible. The CO concentration surges over 6000 ppm once the stream temperature reaches 1600°C, whereas the H2 concentration (not shown) grows from 30 ppm at 307 ms to 80 ppm at the CCOFA elevation. The structures of the other core shells develop from the same transitions with attenuation by significantly lower primary air levels and shorter residence times in the shells from progressively higher elevations. The most important characteristics are collected in Table 6.10. The residence time falls by more than a factor of five, and the temperature becomes cooler by 370°C in moving from the lowest elevation to the highest. But all the SR-values for the gas phase are very close to unity simply because CO2 and H2O—the ultimate products of char oxidation—are most abundant. Whereas all the O2 levels are very low, the CO levels diminish in the shells from progressively higher elevations. The tendency among the NO levels is for more NO from the shells at higher elevations, except for elevation E. Even though the NO levels increase, the extent of char oxidation diminishes significantly for shells from the higher elevations. This trend is due to the smaller portions of primary air that penetrate into the shells from higher elevations, which fall from 36.5% to 21.9% to 20.0% to 15.2% to 20.0% over progressively higher elevations. This variation in conjunction with the shorter residence times at the higher elevations is responsible for the much greater amounts of residual char from the highest elevations. The chemical structure of the innermost shell over the upper furnace also appears in Fig. 6.49. The gas temperature remains at 1600°C for the first second of residence

Table 6.10 Emissions from furnace core shells at the CCOFA elevation. Shell Elev.

τ s

T °C

SR

O2 ppmw

CO ppmw

NO ppmw

XCHAR %

A B C D E

1.042 0.669 0.428 0.308 0.185

1576 1518 1430 1367 1206

0.990 0.970 1.003 1.035 1.026

73 421 407 134 160

6380 1610 410 130 160

87 214 199 625 30

49.5 45.7 33.2 19.6 19.9

ChemNet furnace applications

273

time, then cools continuously from the superheater to the economizer. The SR-value for the gas phase remains near unity for the first 1.5 s then gradually increases in tandem with the surge in the O2 concentration, when char oxidation cannot consume all the entrained O2. The chemical environment becomes much more oxidizing downstream of the superheater inlet. The entrainment fraction in Fig. 6.49 is based on all the O2 in the fuel air, auxiliary air, and CCOFA streams. Whereas the entrainment rate is nearly linear, char oxidation consumes nearly all the added O2 as long as the temperature is close to 1600°C. But when the temperature falls and additional resistances to oxidation come into play, particularly ash encapsulation, the O2 concentration grows from under 1000 ppm to over 2.25% at the furnace exit. The CO level falls continuously during the initial 2.5 s then vanishes before the furnace exit. The shell in Fig. 6.49 makes no contribution to the furnace CO emissions. The NO emissions increase in tandem with the extent of char oxidation throughout the entire upper furnace. Apparently, O2 concentrations over several hundred ppm are sufficient to shut down the NO reduction mechanism seen in the lower furnace. The ultimate NO level from this shell is 470 ppm. The contribution from the burnout of the last half of the char was approximately 380 ppm, because the initial NO concentration was 87 ppm. This implies that if there had been no NO reduction in the lower furnace, the NO emission from this shell would have been greater than 700 ppm. Surprisingly, char continues to burn downstream of the superheater where temperatures have fallen below 1200°C. The ultimate extent of char burnout is 97.6% so the total combustion efficiency is well over 99%. This value was predicted from CBK/E for char oxidation without any contribution from CBK/G for gasification by CO2 and H2O in the lower furnace. The structures of the other core shells develop from the same transitions and are very similar because the entrained air from the quench layer was apportioned on the basis of the residual char in each shell as it entered the CCOFA elevation. The most important characteristics at the furnace exit are collected in Table 6.11. The residence time in the innermost shell is again longer than the rest, but the spread is only 25%. All shells have residence times over 2 s, which should be sufficient for complete burnout provided that O2 is accessible. The shell temperatures are similar across the entire upper furnace. All the SR-values exceed unity because a portion of the excess air fed into the furnace was ultimately entrained into all shells. The O2 levels are approximately 2% at the furnace exit. These levels will be blended with the higher Table 6.11 Emissions from furnace core shells at the furnace exit. Shell Elev.

τ s

T °C

SR

O2 ppmw

CO ppmw

NO ppmw

XCHAR %

A B C D E

3.065 2.858 2.570 2.460 2.348

980 995 984 985 957

1.089 1.036 1.049 1.066 1.048

22,820 16,870 15,340 22,500 15,560

0 0 0 0 0

469 500 501 636 1327

97.6 98.7 98.3 98.5 96.7

274

Process Chemistry of Coal Utilization

level in the quench layer to evaluate furnace flue gas compositions. When O2 is entrained at a rate to completely burnout the char, no CO emissions will be generated by the shells of the furnace core because the gas phase becomes too oxidizing. The tendency among the NO levels is for more NO from the shells from higher injector elevations. The reason is that the shells from the higher elevations have more residual char at the CCOFA elevation (as 100  XCHAR in Table 6.10) and, consequently, more O2 was entrained into these shells. Larger amounts of char bring more char-N into the zone, and a portion of whatever is present will be released as NO. Moreover, the larger O2 entrainment keeps the O2 concentration above the threshold for NO reduction for a longer share of the available residence time, which ultimately enhances the NO emissions. As the entrained O2 was apportioned by the residual char in the shells at the CCOFA elevation, the ultimate extents of burnout are uniformly high. The ChemNet simulations resolve extents of char burnout across the char PSD. At the CCOFA elevation, particles larger than 150 μm are hardly burning, whereas particles smaller than 25 μm have already burned out. At the furnace exit, the entire PSD is ignited, although only particles larger than 100 μm contribute to LOI emissions. The analysis predicts that only the largest portion of the char PSD contributes to LOI, consistent with field test data. The predicted contributions to LOI from the core shells are large char particles, not soot. The quench layer is the source of O2 entrainment into the shells of the furnace core along the upper furnace. The initial gas composition is a mixture of all ribbon products plus all residual fuel air and auxiliary air plus all CCOFA. This stream contains 7.2% O2, 3000 ppm CO, 70 ppm NO, 64 ppm HCN, and 47 ppm NH3. There are no GHCs or H2. A small amount of soot enters the quench layer, constituting under 1% of the total flow. As seen in Fig. 6.50, the adjustments to the emissions levels due to withdrawal of O2 (and H2O and N2) from the quench layer are surprisingly large. The large initial perturbation to the concentrations of O2, CO, NO, and HCN represents the equilibration of the mixture entering the quench layer. Had the instantaneous mixing process to prepare the inlet composition in the calculations been resolved in time, these perturbations would have occurred continuously. Notwithstanding the poor time resolution, the adjustments to the species concentrations should be realistic. The gas temperature increases to 1250°C during the initial 500 ms while the CCOFA streams mix into the furnace flow. Then the gas temperature falls continuously throughout the upper furnace, reflecting heat extraction through the upper waterwalls, superheater, and reheater. The O2 concentration falls continuously while gases are transported into the furnace core shells. After its initial relaxation, the CO concentration gradually grows during the first 1.75 s, then increases at a much faster rate after 2 s. Whereas the HCN concentration changes in tandem with the CO concentration, the NO level diminishes at a uniform rate throughout and is reduced by almost two-thirds from its maximum level across the quench layer. All soot was consumed in less than 450 ms so changes in the emissions are entirely due to reequilibration of chemistry in the gas phase. These results establish three important characteristics: (1) The quench layer is the source of CO in flue gas (in the absence of gross maldistribution and short-circuits).

ChemNet furnace applications

275

8

1300 TGAS

O2

1100

6

1000 900 800

O2, % by mass

7

5 Quench Layer

3000

Quench Layer

150

NO

2500

4 175

CO, ppmw

125 2000 100 1500 75 1000

50

500

25

CO 0 0.0

NO & HCN, ppmw

Gas Temperature, °C

1200

HCN 0.5

1.0 1.5 2.0 Residence Time, s

2.5

0 3.0

Fig. 6.50 (Top) Temperature and O2 and (bottom) (dashed curve) CO, (solid curve) NO, and (dotted curve) HCN levels along the quench layer. The time scale begins at the inlet to the layer immediately above the CCOFA elevation.

It survives the high temperatures and O2 levels because it forms while O2 is withdrawn into the furnace core. (2) NO is reduced in the quench layer, presumably only in the gas phase by CO. (3) Soot does not survive passage through the quench layer and therefore makes no contribution to LOI. All LOI from this furnace is expected to consist of larger residual char particles. Once the quench layer was blended with the products from the five shells of the furnace core and compositions were converted to a volumetric basis, the predicted furnace flue gas emissions are 161 ppm CO; 219 ppm NO; 7 ppm HCN; 1 ppm NH3; 2 ppm NO2, and 12 ppm N2O. At 161 ppm, the CO emissions are six times the reported value but still well below any practical threshold for concern. More important, the ChemNet analysis identifies the quench layer as the source of CO in flue gas and rules out contributions from the furnace core. Nitric oxide dominates the predicted

276

Process Chemistry of Coal Utilization

N-species distribution, as expected. There are non-negligible amounts of N2O and HCN but not NH3 and NO2. The sum of all NOX-species is 233 ppm. At a conversion factor of 733 ppm per lb. NOX/MMBtu, the predicted NOX emission is 0.318 lb. NOX/MMBtu, in agreement with the mean measured value of 0.31. After the residual char levels from the five core shells were averaged, the aggregate extent of char burnout was 98.3%. The total combustion efficiency based on the CO emissions and the residual char emissions is 99.5%. The predicted LOI is 9.3%, which is much higher than the measured value of 0.5%. One possible explanation is that the reducing environment in the furnace core over most of the furnace length promotes substantial extents of char conversion via gasification by steam and CO2. Estimates from CBK/G were based on temperatures of 1400, 1500, and 1600°C for a reaction time of 2.5 s, which is a nominal mean value over all core shells, and nominal concentrations of 22% CO2 and 6% H2O. The predicted extents of char gasification by both agents were 7.0% at 1400°C, 8.5% at 1500°C, and 10.6% at 1600°C. As the gas temperature histories in all core shells except shell E reach maximum values between 1500°C and 1600°C (cf. Fig. 6.46), extents of char gasification probably approach 10%, which is not negligible. The significant contribution estimated for gasification does not rectify the adjustment by 1.6 times to the O2 entrainment history into the core shells noted in Section “Equivalent reactor network”. But accounting for gasification would accelerate the total char conversion rate and, in all likelihood, enable accurate predictions for LOI with the same entrainment profile used in the baseline simulations. Accounting for char gasification will probably perturb the predicted NOX emissions as well. The main reason is that gasification will maintain a highly reducing environment over longer regions of the furnace by generating CO and H2, both of which will rapidly burn away the available O2. Maintaining very low O2 levels with an appreciable amount of CO significantly reduces the NO levels generated during char oxidation (cf. Figs. 6.47 and 6.49). Char gasification enhances both prerequisites for NO reduction. Moreover, a char gasification mechanism will not adversely affect the CO prediction from the furnace core, because the core flows will still contain an abundance of O2 along the convective passes which will still eliminate any residual CO. In summary, the predicted gaseous emissions for the baseline operating conditions are reasonable in that the estimated NOX emission of 0.318 lb. NOX/MMBtu (ca. 233 ppm) matches the mean measured value, and the predicted CO emission is well below the threshold for concern. The analysis identifies the source of the CO emissions as the quench layer moving along the upper furnace walls. Whereas the predicted combustion efficiency is 99.5%, the predicted LOI is many times higher than the reported value. This discrepancy is probably due to the omission of char gasification by CO2 and H2O in the analysis and can be rectified. The complete segregation of volatiles from char near the injectors is a major surprise from the analysis. Volatiles and their associated fuel air and auxiliary air streams move in helical spirals along the walls of the lower furnace, often colliding with adjacent flows. Under the best circumstances, these ribbon flows maintain near-ideal conditions for the conversion of volatile-N species into N2 rather than NO. There is an

ChemNet furnace applications

277

initial surge in the NO level while volatile-N species are released into essentially pure primary air. The SR-value for the gas phase then diminishes while volatiles are released. But it remains three times larger than the whole-coal value because the abundant soot yield does not factor into SR for the gas phase. The early NO is destroyed under weakly reducing conditions while HCN and NH3 levels grow, provided that the entrainment of fuel air and auxiliary air is delayed. A minimum NO level is obtained when all the early NO is reduced away, and all the fixed-N species have been converted into N2. In practice, all the early NO may be eliminated but not all the fixed-N species. So the ultimate NO level from each ribbon stream is determined by the mixing delay. Earlier mixing converts more of the fixed-N species into NO, but it also eliminates more CO and soot. The most favorable situation is most often disrupted by early mixing or short-circuits in the flowfield. Both conditions reduce the time available for N2 production, CO destruction and, especially, soot burnout. Ribbon flows from the higher injector elevations also have much shorter residence times to the CCOFA elevation and are therefore more likely to retain more noxious gases. For all these reasons, the analysis predicted huge variations among the levels of O2, NO, CO, fixed-N species, and soot at the CCOFA elevation from particular fuel injectors. There were clear tendencies for lower levels of NO and O2 and higher levels of HCN, NH3, and soot in the ribbon products from progressively higher elevations. However, higher HCN and NH3 levels are associated with higher ultimate NO emissions, because the unavoidable conversion of fixed-N species in the quench layer generates more NO. The predictions also displayed variations among the emissions from ribbons from the injectors on the same firing elevation, due to erratic collision patterns among adjacent flows. Such predicted variability is based on a CFD simulation with uniform coal injection rates at all elevations with variable air flowrates. Clearly, uniform injector settings do not ensure the most uniform combustion behavior due to the interactions among the flows from adjacent injectors, albeit within the fidelity of 3D furnace CFD. It remains to be determined whether or not injector fine-tuning can enhance the ribbon combustion efficiency, although the ChemNet analysis predicts a large potential payoff. Indeed, ribbon products are the dominant sources of NO and CO into the upper furnace. Shells of the furnace core sustain char oxidation but no volatiles combustion or soot chemistry. Nevertheless, gas-phase chemistry is crucial to the emissions from this region. All the O2 fed into the shells with the char is consumed very rapidly, often within 100 ms (cf. Fig. 6.47). Additional burnout is limited by the entrainment rate of fuel air and auxiliary air, which is very slow. Hence, the local chemical environment remains reducing throughout almost all the lower furnace; in fact, O2 concentrations stay under the threshold that disrupts NO reduction, which is a few hundred ppm. Consequently, much less NO is generated in core shells across the lower furnace than expected from isolated char particles in a typical oxidizing furnace environment. The variations in emissions from the different shells are significant because there is less O2 penetration and shorter residence times for shells from progressively higher injector elevations.

278

Process Chemistry of Coal Utilization

The entrainment of O2, H2O, and N2 from the quench layer across the upper furnace is much faster. Faster entrainment accelerates the char burning rate, but it also maintains O2 levels above the threshold for NO reduction. Consequently, NO emissions from the upper furnace are much more consistent with the assigned value for the char-N conversion factor. In total, the core shells contribute 56% of the flue gas NO emissions. This contribution is considerably smaller than expected, due to the extensive NO reduction across the core of the lower furnace. The predicted combustion efficiency was 99.5%, although the predicted LOI was still too high. However, according to CBK/G, about 10% of the char is probably converted via gasification by steam and CO2 in this furnace. Adding a char gasification mechanism is expected to rectify the problems with the LOI prediction and also strengthen the basis for the NOX prediction. Chemical adjustments to the quench layer composition are surprisingly strong. Where the temperature stays hotter than 1000°C, soot burns out quickly, the CO level increases very gradually, and the NO level is more than halved. The quench layer analysis indicates that there is no soot in LOI; that the quench layer is the only source of CO in furnace flue gas; and the quench layer mediates the NOX generated in the ribbon products to 44% of the ultimate furnace NOX emissions.

6.4

Commercial CFBC furnaces

As for pc furnaces, a diverse assortment of computer simulations has been developed for circulating fluidized bed combustors (CFBCs), including species mass balances in 1D with empirical hydrodynamics; population balances for fuel, ash, and sorbent particles; CFD; and large eddy simulations of particle clusters entrained through risers. The treatments for the hydrodynamics and particle dynamics cover an enormous range of mathematical sophistication, but the chemistry submodels are consistently rudimentary: The ultimate yields and compositions of volatiles are specified as input parameters; volatiles combustion is treated with a few global, nth-order reactions; and char burning rates are based on one global nth order reaction in parallel with film diffusion of O2 to the external particle surface, and shrinking core behavior. The analysis in this section was the first to include soot production and combustion. The ChemNet CFBC analysis combines the comprehensive reaction mechanisms from Chapter 5 with hydrodynamics and particle dynamics that were deliberately simplified to accommodate detailed chemistry (Niksa et al., 2017). The main objectives are to identify (i) which fuel components make the largest contributions to UBC in flyash and (ii) the factors responsible for incomplete conversion of diverse coals. As seen below, the predicted fuel conversion efficiencies are consistent with reported flyash LOI from numerous full-scale CFBCs for diverse coals. The validation database covers virtually the entire operating domain for commercial CFBCs with 20 test cases in 11 full-scale CFBCs whose range of coal quality extends from brown coals to anthracite. The predicted fuel burnout histories reveal where in a CFBC different fuel components burn and which ones determine LOI levels, including heretofore unrecognized roles for heterogeneous ignition and thermal annealing of char.

ChemNet furnace applications

279

In principle, the chemistry package predicts NOX emissions as demonstrated in the cases of pc furnaces. However, NOX emissions from CFBCs are governed by detailed chemistry as well as rates of mixing of primary and secondary air into the process stream as it moves through the splash zone and along the riser. Unfortunately, the CFBC literature offers little guidance on the flow patterns in splash zones, and CFD simulations to specify air entrainment rates were not available for this analysis. In lieu of objective estimates for the mixing rates, the analysis invokes the limiting case of instantaneous mixing of bubbles, emulsion gas, and air streams in the splash zone, as a means to avoid unbounded tuning of mixing parameters. Unfortunately, this assumption undermines the predicted NOX and CO emissions, which will not be considered further. CFBCs consist of a dense bottom bed, splash zone, riser, exit section, cyclone, and external circulation return. Dense bottom beds are fed by primary air, coal, recirculated bed ash, and, usually, limestone to control SO2 emissions. Within the dense bed, the flows of primary air and volatiles partition into bubbles and an emulsion phase that contains nearly all solids. Most of the coal grind releases its volatiles within the bed, which are then converted by secondary volatiles chemistry into noncondensable fuel mixtures (CO, CH4, C2H2, HCN, H2, H2S) plus soot. The residual char is ground by mechanical attrition into fines and a smaller PSD of coarse char. So the fuel components with distinctive burning rates are noncondensable gas mixtures, soot, char fines, and char. The splash zone is fed by elutriated particles from the bottom bed, especially the particle clusters ejected by bubbles that happen to rupture through the bed surface. Depending on the distributions of size, density, and terminal particle velocities, particles either fall back into the bottom bed or form a dispersed suspension that moves upward in a core flow through the riser. The riser flow is usually supplemented with secondary air injection at multiple elevations. It partitions into a dense, downward, annular particle flow along the walls and a dilute upward core flow. The particle concentrations in the core and wall layers decay exponentially with height, so solids must move into the dense wall layer along the riser. Consequently, the area of the core expands with height while that of the particle layer becomes thinner. At the exit zone, a portion of the particles recirculate internally into the downward wall layer while the rest are accelerated through the exit duct into cyclones. The internal recirculation rate is a minor contribution in full-scale systems but a major one in lab-scale CFBCs with high aspect ratios and very high particle concentrations at the top of the riser. The cyclone rejects flue gas from the CFBC along with unconverted char and flyash. It returns the larger and denser particles into a downcomer, where they move slowly downward into the particle seal and, ultimately, into the dense bed return. The CFBC analysis proceeds through three independent stages. First, particle dynamics calculations determine the partitioning of coal and sorbent into fines, carryover, and ejection streams from the dense bottom bed. The ejection of small portions of the wakes on bubbles is the only transport mechanism that can accommodate reported ash circulation rates. In the splash zone, particle slip velocities determine the cutoff between particles that fall back into the dense bed and those that move upward into the riser. Gas velocities diminish along the riser due to thinning of the

280

Process Chemistry of Coal Utilization

dense particle wall layer with elevation and to cooling along suspended heat transfer surfaces but accelerate via injection of secondary air. Secondary air injection morethan-compensates for the decelerating factors, so all combustibles are assumed to remain suspended while they move upward along the riser. Nearly all char passes through the cyclone into the gas cleaning system due to its relatively very low density and diminished particle size. The second ChemNet calculation pass assigns extents of combustion for noncondensable fuels, soot, char fines, and char based on rapid mixing of various streams in one CSTR for the splash zone in series with 40 CSTRs for the riser and exit zone. Bubbles and emulsion gas are completely mixed by the exit of the splash zone, and this flow then accepts secondary air along the riser based on a mixing law for jets in crossflow. Oxygen is apportioned to the various noncondensable and solid fuels based on the competitive chemical reaction mechanisms for the different fuel components. The third calculation pass determines extents of Ca-utilization by limestone sorbents and the SO2 capture efficiency. The first and third calculation passes are beyond the scope of this chapter, although the dense bed analysis (Niksa et al., 2017) and the predicted SO2 capture efficiencies (Niksa et al., 2015) have been reported. The next section describes the second calculation pass.

6.4.1 Splash zone, riser, and exit zone Within a splash zone, the ejected particle fluxes partition into coarse particle clouds that settle back into the dense bed and suspensions of smaller particles that have sufficient buoyancy to move upward through the riser, based on terminal velocities across the solids PSDs. Air released from bubbles in the dense bed mixes with the fuel mixtures from the bed emulsion and burns, starting with the most reactive fuel components. The burnout sequence for the various fuel components is determined by their distinctive oxidation kinetics. Unfortunately, there is little, if any, quantitative information on bulk flow patterns and rates of mixing in splash zones in commercial CFBCs. In the simulations, mixing of primary air in the splash zone is represented with a single CSTR, while secondary air injections into the splash and along the riser are represented with finite-rate decaying entrainment for jets in crossflow into a series of CSTRs that closely approximate plug flow. The gas velocity is determined from the superficial gas velocity from the top of the dense bed, with adjustments for the diminished flow area and the appreciable volume fraction of solids. The nominal transit time across the splash zone for each increment in the solids PSD is given by the ratio of zone height to particle slip velocity, which depends on the particle size and particle density. The height of the splash zone above the dense bed is estimated as twice the bubble diameter through the top of the dense bed. Since the slip velocities through the splash and riser vary with particle size, each size increment passes through the CSTR network in distinctive transit times, and these variations are substantial enough to affect extents of char conversion, as illustrated below. A multitude of fuel species—CO, H2, CH4, C2H2, HCN, H2S, soot, char fines, and coarse char—compete for the available O2 along the splash zone and riser. Realistic

ChemNet furnace applications

281

chemical kinetics for each distinctive combustion process govern the competition for O2 without any a priori assumptions on this partitioning. For the ChemNet reaction mechanisms in Chapter 5 and our focus on LOI predictions, the only adjustable parameter in the analysis is the initial char oxidation reactivity. This parameter was assigned in calibration for LOI for any coal type from commercial CFBCs, as presented below in Fig. 6.51. The primary process stream in this reaction system begins as the gases in the bubbles and emulsion from the dense bottom bed move into the CSTR for the splash zone. When a portion of the coal feed remains in the splash zone or penetrates into the riser, succeeding reactors blend secondary volatiles into the process stream, along with any conversion products from soot, char fines, and char. Entrainment increments of secondary air are also added to individual CSTRs. All these additions are evaluated from the FLASHCHAIN®-based kinetics for devolatilization and tar decomposition for noncondensable fuels; from the NSC soot burning rates for soot; from char burning rates for conversion products from char fines and char from CBK/E; and from the mixing submodel for secondary air. Similarly, the temperature of each reactor in the series is based on a measured temperature profile across the riser. Mean gas residence times in each reactor are based on the calculated gas velocities and the elevation increment that each reactor represents. In the CFBC chemistry simulations, the splash zone is represented with a single CSTR; the riser by 20 reactors in series; and the exit turn and exit duct by another 20 reactors. Due to the large number of reactors used for the riser, the flowfield is essentially in plug flow. All of these specifications could be refined by CFD simulations. The species balance for each CSTR is the same as those for furnace streams in Chapter 5 except for multiple additional terms for char fines and entrained char particles, according to  vC, i,O2 Mi ð1  xASH Þ  ηC,O F0C + ηC,C FC, C + ηC,EJ FC,EJ + ’ MC   ð vC,i,O2 yC, f Ff Mi ð1  xASH ÞXC af vS,i,O2 F∞ S Mi X S +  ωi Mi dV MC’ MC’ (6.3)

+1 FjP+ 1 yjP,i  FjP yjP,i  FjE yjE,i ¼

where FjP and Fj+1 P are the mass flowrates into and out of the jth CSTR, respectively, j and FjE is the increment of secondary air; yjP, i, yj+1 P, i, and yE, i are the corresponding mass 0 fractions of species i in the respective streams; FC, FC, C, and FC, EF are the char flowrates from the coal feedstream, and the carryover and ejected streams from the dense bed; ηC, O is the extent of char conversion integrated across the entire char PSD; ηC, C and ηC, EJ are analogous contributions for the carryover and ejected char streams; XC(af) is the increment in the extent of char conversion for char fines of size af; vC,i,O2 is the stoichiometric coefficient for species i in the char conversion reaction; Mi, MS0 , and MC0 are the molecular weight of species i and the mean element weights in soot and char; XS is incremental soot conversion; vS,i,O2 is the stoichiometric

282

Process Chemistry of Coal Utilization

coefficient for species i in the soot conversion reaction; ωi is the net volumetric generation rate of species i from homogeneous chemistry; and dV is a volume increment. As the composition within the jth CSTR is uniform, by definition, the last integral is readily evaluated as the product of the mean homogeneous reaction rate and the reactor volume weighted by the RTD. Note the explicit incorporation of results from the particle dynamics calculations for the dense bottom bed, which specify the partitioning of char into carryover and ejection streams and also determine the flowrate of fines from mechanical elutriation in the dense bed. The overall mass balance has a similar form, except that no homogeneous reaction term appears because chemistry in the gas phase cannot change the mass of the process stream. This system of equations comprises two implicit equations for two unknowns j+1 for each species, Fj+1 P , and yP, i. Once specified, these unknowns determine the extents of solids conversion, as char oxidation kinetics may be evaluated once the gas composition has been specified; they similarly determine XS. The specified unknowns also determine the homogeneous reaction rate, ωi, because the homogeneous reaction mechanism specifies all production rates in terms of the species concentrations. Independent balances on the three solids streams for char, char fines, and soot have a similar form, without the contributions for air entrainment and homogeneous chemistry. They determine the flowrate and PSD of the char effluent and the effluent flowrates of char fines and soot.

6.4.2 Interpreting LOI emissions from commercial CFBCs The CFBC validation database contains eleven CFBCs with ratings from 1.2 MWth to 235 MW. Tests were replicated at five units, usually with different fuels or at different loads, so there are 20 distinct case studies. Commercial CFBCs in the small size class, from 20 to 35 MW, and for power generation, from 100 to 300 MW, are well represented. The flow cross-sections and furnace heights were specified for all CFBCs, but only two heights of the dense bottom beds were reported. The larger CFBCs have multiple fuel injection ports, and elevations for fuel injectors and solids return ports were usually indicated. Secondary air elevations were reported for all cases that had secondary air injectors except one. Dense bed operating temperatures were between 830°C and 900°C, except for the 960°C imposed at one CFBC. The CFBCs without limestone injection tend toward hotter operating temperatures. The maximum temperatures were somewhat hotter and usually recorded at about one-third of the riser height. Temperatures drop across the upper riser elevations, usually by more than 50 degrees. Nineteen different coals cover ranks from brown coals to anthracite with especially good coverage of brown coal and hv bituminous samples. Many samples have excessive ash contents of 40% to 50%, particularly in the Chinese CFBCs. All specifications were taken from the field test reports and implemented directly in the simulations. Fuel feedrates were assigned to match the reported maximum furnace rating, based on a thermal conversion efficiency of 33.3%. Excess air levels were either reported or assumed to be 20%. The primary air split was either reported or assumed to be 60%. PSDs for coal were either reported or assumed to give a mean size from 2 to 3 mm. PSDs for limestone were either reported or assumed to give a

ChemNet furnace applications

283

40 35

LOI, wt. %

30 25 20 15 10 5 0

0

10

20

30

40

50

Coal Index I, g/MJ

Fig. 6.51 Assigned (●) flyash LOI for all CFBCs in the validation database based on the curve reported for commercial Chinese CFBCs by Yang et al. (2005). (Reproduced with permission from Niksa S, Sakurai Y, Fujiwara N. Predicting the conversion efficiencies of any coal type in CFBCs, Energy Fuels 2017b;31:4507–19. The American Chemical Society.)

mean size of 150 μm. Each simulation took under 10 min on a microprocessor operating at 3.2 GHz. There are no adjustable parameters in the reaction mechanisms for devolatilization, secondary volatiles pyrolysis, volatiles combustion, and soot burnout. A single reactivity parameter for char burnout was specified for each coal to match a correlation of the reported LOI levels from two dozen commercial Chinese CFBCs (Yang et al., 2005), as seen in Fig. 6.51. The X-axis is a volatility index defined as the ratio of the proximate volatile matter, in daf wt%, to the lower calorific value, in MJ/g, which could be evaluated for all coals in the database from standard coal properties. This correlation represents coals across the entire rank spectrum whose LOI levels varied from 4 to 28 wt%. In so far as the Chinese LOI correlation pertains to the CFBCs in the validation database, this calibration procedure ensures that the coal conversion efficiencies from the simulations are reasonably accurate. It was also reassuring that none of the adjustments to the default reactivities from CBK/E were inordinate; among 19 different coals in the database, the median adjustment to the results for default char burnout kinetics was 20%. Moreover, three cases for a Chinese CFBC demonstrated that hotter furnace temperatures and greater furnace loads reduced predicted LOI, in accord with established trends (Zhao et al., 2005). Throughout this section, generic results are based on a commercial furnace rated at 105 MWth that fires a blend of hv bituminous coals; it is labeled as CFBC E. The elevation of the bottom of the exit duct is 20 m above the distributor plate, and the riser

Process Chemistry of Coal Utilization

Case E

TGAS

2.5

Transit Time

880

1.5

Char Fines Soot

890

2.0

Gas 870 Flow

1.0

100

900

Char

60

40

20

0.5 860 0.0

80

0

5 10 15 Elevation Above Distributor, m

20

BO for Char, Fines & Soot, %

3.0

Gas Temperature, °C

Transit Time, s & Fractional Gas Flow

284

Case E 0

5 10 15 Elevation Above Distributor, m

20

0

Fig. 6.52 (Left) Operating conditions for CFBC E and (right) predicted burnout profiles for soot, char fines, and char. (Reproduced with permission from Niksa S, Sakurai Y, Fujiwara N. Predicting the conversion efficiencies of any coal type in CFBCs, Energy Fuels 2017b;31:4507–19. The American Chemical Society.)

includes two secondary air injectors at 2.3 and 6.9 m. The splash zone extends from 1 to 2.2 m. Profiles of transit time, temperature, and flowrate of the process stream appear in Fig. 6.52. The gas flowrate is normalized by the value at the cyclone exit. The increases in the flowrate reflect both secondary air additions as well as the conversion of solid fuels to gaseous products. Due primarily to the secondary air injections, the transit time per unit elevation diminishes across the splash zone and the first several meters of the riser. These additions plus the production of gaseous products more than double the gas flow across this region. Thereafter, the gas flowrate is almost steady because only larger char remains to be converted, and transit time increases in direct proportion to elevation. Midway up the riser, the gas temperature passes through a weak maximum and starts to diminish when the flow contacts a pendant superheater from 9.2 to 12.7 m. Due to the assumption in the simulations that emulsion gases and bubbles are fully mixed within the splash zone (which covers elevations from 1.0 to 2.2 m in Fig. 6.52), most of the O2 from the primary air stream and all noncondensable fuels and intermediates were completely consumed in the splash zone in Fig. 6.52 and in all other CFBCs in the database, which is too fast. This flaw reflects the absence of any objective means to estimate the mixing rates in this section of the CFBC and was adopted to avoid unbounded tuning of mixing parameters. It should eventually be replaced by a finite-rate mixing process tuned to CFD simulations. As seen in Fig. 6.52, char fines burn fastest among the solid fuels, by far, and are completely consumed in the splash zone with most coal types. The relatively very fast burning rate of fines reflects the high intrinsic reactivity in conjunction with the very small size, which eliminates all transport resistances. Conversely, hardly any soot or large char burns in the splash zone in this particular case. Soot burns out through the first three-fourths of the CFBC, while coarse char particles cannot completely burn out in the available transit time. The important implication is that char fines can

ChemNet furnace applications

285

Exit

Exit

Case E

15.1

80 Char Burnout, %

20.0 m

8

20.0 m 15.1 9.2 5.7

60

Splash 9.2 40

7 6 5 4

5.7

3

Transit Time, s

100

2

20

1 Case E

Splash

0

0 0

100

200

300

400

Initial Coal Diameter, mm

500

600

0

100

200

300

400

500

600

Initial Coal Diameter, mm

Fig. 6.53 (Left) Extent of char burnout vs. initial coal particle size from the splash zone through the exit duct; and (right) Char transit times to the indicated elevations along CFBC E. (Reproduced with permission from Niksa S, Sakurai Y, Fujiwara N. Predicting the conversion efficiencies of any coal type in CFBCs. Energy Fuels 2017b;31:4507–19. The American Chemical Society.)

effectively compete for O2 with noncondensable fuels. It is the only particulate form that can mediate the burning rates of noncondensable fuels, and thereby influence N-species conversion (although soot may catalyze radical recombinations and heterogeneous NO reduction). However, this implication cannot universally apply to all coal types, because intrinsic char oxidation rates strongly diminish for coals of progressively higher rank (Niksa et al., 2003). Over the validation database, LOI was entirely determined by incomplete burnout of large char particles in most cases, with no contribution from char fines. But in six cases, incomplete conversion of fines was a contributing factor, albeit a small one. Only anthracite produced fines that burned slower than coarse char to give burnout levels of 82% for fines vs. 96% for coarse char. In pc furnaces, extents of char burnout monotonically decrease for progressively greater char sizes (cf. Figs. 6.24 and 6.26), and the UBC in flyash LOI exclusively comes from the largest particles in the coal grind. But with the much coarser coal grinds fed into CFBCs, the size dependence is more complex, as seen in Fig. 6.53 (and also in Figs. 6.54 and 6.55, below). Extents of burnout do diminish with increasing size for the smallest particles in the entrained char PSD, but the burnout then grows for even larger sizes. The minimum reflects the much longer transit times for the largest particles, which compensate for the stiffer transport resistances in their burning rates, and explains why the burnout for the largest particles exceeds that for moderately smaller ones. Consequently, the largest sizes in the char PSD do not make the largest contributions to LOI. Rather, LOI comprises a fairly broad segment of the larger char particles whose contributions by size are variable. Among entrained char particles (excluding fines), only the smallest sizes effectively compete for O2 in the splash zone. This tendency is illustrated in Fig. 6.54 by histories of burnout and particle temperatures for three sizes that span the PSD of entrained char. The smallest size is rapidly consumed in the splash zone and lowest

286

Process Chemistry of Coal Utilization 1400

1.0 560

297 mm 560

Fractional BO

0.8

1200

0.6 1000 0.4 34

800

Particle Temperature, °C

297 mm

34

0.2 Case E 0.0

0

2

4 6 Transit Time, s

8

0

2

4 6 Transit Time, s

8

600

Fig. 6.54 (Left) Fractional char burnout vs. transit time for three sizes of coal particles from the outlet of the splash zone through the exit duct; and (right) particle temperatures during these combustion histories along CFBC E. (Reproduced with permission from Niksa S, Sakurai Y, Fujiwara N. Predicting the conversion efficiencies of any coal type in CFBCs. Energy Fuels 2017b;31:4507–19. The American Chemical Society.)

1

1.0

0.6

1000

h

0.2

Chi

900

1.0

1.5 2.0 Transit Time, s

2.5

0.6

0.4

k/k0

0.01

Case E 0.5

0.8

0.1

1100

TP

0.4

1200

Chi

d/d0 & h

d/d0

Particle Temperature, °C k/k0

0.8

0.0 0.0

1.0

1300

0.2 Case E

800 3.0

1E-3 0.0

0.5

1.0

1.5 2.0 Transit Time, s

2.5

0.0 3.0

Fig. 6.55 (Left) Combustion history for char from 297 μm coal particles including particle temperature, scaled particle size, and effectiveness factor, η; and (right) the associated decay in the char oxidation reactivity and the Chi parameter. (Reproduced with permission from Niksa S, Sakurai Y, Fujiwara N. Predicting the conversion efficiencies of any coal type in CFBCs. Energy Fuels 2017b; 31:4507–19. The American Chemical Society.)

riser elevations, even though this size class did not achieve a fully ignited burning state. Even though it burned under chemical control, the intrinsic oxidation reactivity was still fast enough to complete burnout. In contrast, the two larger sizes ignite at temperatures that are much hotter than local gas temperatures. At face value, the abrupt and simultaneous surges in both burnout and temperature look like the signatures for ignition delays. Actually, the times before the particles ignite are the estimated transit times through the splash zone, which vary from 400 ms to 4.5 s for

ChemNet furnace applications

287

this char PSD. As only primary air moves through the splash in this case and as nearly all the noncondensable fuels and char fines are converted in the splash, the estimated O2 concentration is less than 0.2%, which is too low to ignite coarse char. Only particles smaller than 100 μm give appreciable burnout in the splash zone. Larger char particles ignite as soon as they contact the much greater O2 levels above the secondary air injectors, as evident in Fig. 6.54 by much faster burning rates and abrupt surges of 300 to 400°C in the particle temperatures. At these hotter temperatures, most char particles burn out within their size-dependent transit times, at which point the particle temperatures relax back to the gas temperature. As the particle temperatures after ignition exceed the local gas temperatures by hundreds of degrees, coarse char is subject to thermal annealing. A complete burnout history and the impact of thermal annealing on the intrinsic reactivity appear in Fig. 6.55. This combustion history includes the decay in the normalized char particle diameter, which indicates continuous shrinkage during char burnout, although the particle density also diminished throughout this burnout history. Consistency with shrinking core behavior is gauged by the Chi-parameter in the right panel, which is the ratio of the instantaneous burning rate to the burning rate determined by film diffusion of O2 onto the external particle surface. Chi ranges from zero to unity and thereby indicates the transition from chemical kinetic control to film diffusion control. It approaches unity as the char moves across the splash zone because burning rates become limited by film diffusion for progressively lower O2 levels, and O2 levels are very low in this splash zone. After ignition, this parameter tracks the variations in particle temperature because the surface oxidation kinetics have a much stronger temperature dependence than the O2 film diffusion rate. Accordingly, Chi exceeds 0.8 soon after ignition and then varies from 0.8 to 0.9 during char burnout. Such large values indicate substantial mediation of the char burning rate by film transport, with an appreciable contribution from the oxidation kinetics. Strictly speaking, shrinking core behavior would be associated with a Chi-value of unity throughout burnout. The effectiveness factor, η, in the left panel of Fig. 6.55 also does not corroborate shrinking core behavior. This parameter is a gauge for the penetration of O2 toward the centers of char particles, where unity indicates complete penetration. It is very low at ignition, but rises continuously during burnout to a value of two-thirds, reflecting the expansion of the internal porosity throughout char conversion. The final important aspect of char conversion, in this case, is the impact of thermal annealing on the intrinsic char oxidation reactivity. This feature is illustrated in Fig. 6.55 by the ratio of the instantaneous rate constant for oxidation to its initial value. Even while the particle traverses the splash zone at 880°C, the reactivity diminishes by a factor of fifty. As the annealing rate surges after ignition due to the much hotter temperatures, the reactivity decays at a faster rate by almost another order of magnitude. The simulation results only account for annealing after char has been expelled from the dense bed, so the initial impact would have been slightly greater if soaking in the dense bed had been included. This strong impact of annealing affects the ignition characteristics and, in principle, might also reduce the potential to manage LOI by recirculating the largest unconverted char particles through multiple passes through the CFBC. To investigate

288

Process Chemistry of Coal Utilization

this possibility, char particles of the density and size at the end of a first pass through the CFBC were artificially annealed to the same reduction in burning rate before they completed another identical cycle through the CFBC. Due to their much lower particle density and smaller size, these particles ignited and burned slightly hotter on the second pass, despite their much slower reactivity at the ignition point, and achieved complete burnout. This calculation did not account for fragmentation during the second exposure to the dense bed or for the shorter transit time of a lighter, smaller particle through the second pass, which would reduce the predicted extent of burnout at the CFBC exit. Even so, it gives no evidence that thermal annealing would retard ignition during subsequent passes through a CFBC because most annealing occurs upstream of the ignition point in the first pass.

6.4.3 Coal quality impacts Two cases in this section illustrate the burnout characteristics for coals at the extremes of the rank spectrum. The CFBC in Case J burns brown coal with 30 wt% ash, as received, which generates a char whose ash level exceeds 60%. The initial oxidation reactivity is 50 times faster than the char from the bituminous blend considered in Figs. 6.52–6.55. Secondary air is injected into the dense bottom bed and also above the splash zone, so this splash zone has much greater O2 levels than many other CFBCs. Burnout as a function of the initial coal size is plotted for several elevations in Fig. 6.56, along with histories of burnout and particle temperature for two size classes. This case depicts the expected tendency for pc furnaces for monotonically decreasing burnout for progressively larger particle sizes. But the multiple reactions and transport mechanisms, in this case, expose complexity across the entire PSD. The initial reactivity is fast enough for the entire PSD to compete with noncondensable fuels and char 1150 100

Exit 20 m 15.9 9.2 5.0 Splash

60

40

20

1100

115 µm

80 Char Burnout, %

Char Burnout, %

80

BO BO

1050

60

1000 950

40

340 µm

20

TP

TP 900

0

0

850

Case J

Case J 100

200 300 Initial Coal Diameter, µm

400

0

0

2

4 Transit Time, s

6

Particle Temperature, °C

100

8

800

Fig. 6.56 (Left) The extent of char burnout vs. initial coal particle size from the splash zone through the exit duct; and (right) Char burnout and particle temperature histories for 115 and 340 μm coal along CFBC J which fires brown coal. (Reproduced with permission from Niksa S, Sakurai Y, Fujiwara N. Predicting the conversion efficiencies of any coal type in CFBCs. Energy Fuels 2017b;31:4507–19. The American Chemical Society.)

ChemNet furnace applications

289

fines for the available O2 in the splash zone. Indeed, burnout in the splash varies from 40% to 97%, with a very weak minimum in the middle of the PSD. The important implication of these high conversion levels is that transit times through the riser are much less sensitive to size, and therefore much more uniform than the situation in Fig. 6.53 because particle velocities across the PSD relax to the gas velocity. Even though the entire PSD ignites and reaches particle temperatures well above the local gas temperature in the splash zone, two factors immediately extinguish the combustion into a much cooler slow-burn state: (1) the large diffusional resistance across the ash encapsulation layer surrounding the combustibles; and (2) thermal annealing. Thereafter, the particles burn isothermally at a slower rate and achieve ultimate extents of burnout determined by their relatively uniform transit times through the riser. This is why the curves of burnout vs. size in Fig. 6.56 are similar for all elevations. Hence, with low-rank coals, thermal annealing and, for high-ash coals, the resistance of an ash encapsulation layer, are the determining factors on LOI levels. The opposite extreme is based on the CFBC in Case A4, which burns anthracite with 8% ash and one-fourth the oxidation reactivity of the bituminous blend in Figs. 6.52–6.55. This furnace uses a much lower staging level than the others, so the estimated O2 concentration in the splash zone is much greater than the other cases. As seen in Fig. 6.57, this case gives the most complex size dependence of all in the predicted extents of burnout; at each elevation burnout as a function of size first passes through a local maximum before it diminishes through a local minimum at intermediate sizes, and then rises again for the largest size classes. The segments from the local maximum through the largest sizes reiterate the impact of longer transit times for progressively larger particles (cf. Fig. 6.53). The segment from the smallest size to the local maximum reflects ignition problems for the smaller sizes. As seen in the burnout histories for the extreme size classes in Fig. 6.57, smaller sizes never achieve a fully ignited state due to their relatively faster heat loss rates to the entrainment stream,

80 Exit 10.3 m 60

BO

80

7.6 5.1

40

TP

3.0

1100

1000 40 900

60 mm 20

Splash Case A4

0

760 mm

0

1200

60

1.8 20

BO

Particle Temperature, °C

100

Char Burnout, %

Char Burnout, %

100

150 300 450 600 Initial Coal Diameter, mm

750

0

Case A4

TP 0

1

2

3 4 Transit Time, s

5

800

6

Fig. 6.57 (Left) The extent of char burnout vs. initial coal particle size from the splash zone through the exit duct; and (right) Char burnout and particle temperature histories for 115 and 340 μm coal along CFBC A4 which fires anthracite. (Reproduced with permission from Niksa S, Sakurai Y, Fujiwara N. Predicting the conversion efficiencies of any coal type in CFBCs. Energy Fuels 2017b;31:4507–19. The American Chemical Society.)

290

Process Chemistry of Coal Utilization

whereas larger sizes ignite to particle temperatures that remain much hotter than the local gas temperature throughout most of the transit time. Consequently, the smaller particles burn slower than the larger particles and thereby have lower extents of burnout. In fact, they burn with Chi-values approaching zero and effectiveness factors approaching unity, which is the opposite limit from shrinking core behavior. This behavior also explains why char fines only contribute to LOI for coals with the slowest oxidation reactivity. Beyond the local maximum, extents of burnout diminish for larger particles because their film diffusion rates are slower, and also because thermal annealing partially extinguishes the fully ignited state sooner for progressively larger particles. Eventually, this tendency is compensated by the longer transit times for larger particles, which enhances burnout. Consequently, across the PSD for low reactivity chars, burning rates shift from chemical reaction control (Chi ¼ 0; η ¼ 1) toward film diffusion-limited burning mediated by internal pore diffusion (Chi > 0.5; η < 0.2). No size class exhibits shrinking core behavior. In summary, the primary functions of dense bottom beds are to partition the fuel into fines and coarse particles and the sorbent into fines, flyash, and circulating sorbent; and to eject circulating ash into the riser. Most of the CFBCs passed more char fines than char particles into the risers, and the larger char particles had much smaller sizes than the mean sizes of the coal grinds. Dense bottom beds should therefore be recognized as highly effective grinding devices that micronize most of the coal feed before it is subjected to char burnout. Due to the assumption in the simulations that emulsion gases and bubbles are fully mixed before they burn in the splash zone, most of the O2 from the primary air stream and all noncondensable fuels (CH4, C2H2, H2, HCN, H2S) and intermediates (CO, H2) burned out in the splash zone. Among the solid fuels, char fines burn fastest, by far, and are completely consumed in the splash zone with most coals. Conversely, hardly any soot burns in the splash zone due to its low intrinsic reactivity. Extents of coarse char burnout in the splash zone are also determined by the intrinsic oxidation reactivity, which spans two orders of magnitude for coals across the rank spectrum. With reactive low-rank coals the entire char PSD can be at least partially converted, whereas with much less reactive bituminous coals only the smallest char sizes can compete for the available O2. With the least reactive low volatility coals, ignition may become problematic. Most important, the rank dependence for char oxidation reactivity determines the mode of burning which, in turn, determines which portions of the entrained char PSD make the greatest contribution to flyash LOI. With the most reactive coals, all sizes ignite in quasi-steady combustion at temperatures several hundred degrees above local gas temperatures. These burning rates are mostly governed by film diffusion with appreciable mediation by internal pore diffusion. The main mitigating factors are (i) thermal annealing, which can partially extinguish the fully ignited state; and (ii) for chars with high ash loadings, the diffusional resistance through ash encapsulation layers. When both mitigating factors come into play, the largest char particles make the greatest contributions to LOI. With coals of intermediate reactivity, the smallest sizes in the char PSD never ignite to a fully developed state, but their chemically

ChemNet furnace applications

291

limited burning rates are still sufficient to consume all their combustibles. Even though the larger sizes will fully ignite, the strong size variation in transit times across a riser compensates for the size dependence in the burning rates, so that intermediate sizes in the PSD make disproportionate contributions to LOI. With the least reactive chars, the smallest sizes never ignite, and their intrinsic burning rates are too slow to completely burn out the combustibles. So the smallest char and even char fines contribute more to LOI than the largest coarse char. In fact, except for the largest particles, the entire char PSD from the least reactive coals contributes to LOI. The unifying phenomenon for LOI is the potential for ignition to a near-diffusionlimited burning rate across the char PSD. LOI levels increase whenever the ignition is suppressed in any portion of the char PSD and when fully ignited states are extinguished as char moves through the riser. Ignition is suppressed by low intrinsic oxidation reactivities starting from the smallest char sizes. Fully ignited particles may be extinguished by thermal annealing with any coal and by ash encapsulation with coals with excessive ash levels. Thermal soaking at 850°C for a few seconds diminishes char reactivity by nearly two orders of magnitude due to thermal annealing, and the annealing rate surges after ignition due to the much hotter char particle temperatures. But thermal annealing does not seem to undermine the potential to manage LOI by recirculating the largest unconverted char particles through multiple passes through the CFBC because the annealing occurs upstream of the riser even on the first pass. Char oxidation mechanisms that impose shrinking core behavior across the char PSD cannot possibly depict the required range of behavior to accurately predict LOI levels from diverse coals. Rather, the burning mechanism must automatically shift from reaction control through mediation by pore diffusion to film diffusionlimited burning, and also factor in thermal annealing and the transport resistance for ash encapsulation.

6.5

Summary of ChemNet for furnaces

For coal-fired furnaces, ChemNet CFD postprocessing provides an alternative to conventional CFD postprocessing whereby rudimentary N-conversion schemes are applied to the primary CFD temperature and concentration fields to estimate furnace NOX emissions. One key difference is that in ChemNet, CFD simulations only specify temperature histories and mixing rates, not the species concentration fields that directly reflect the rudimentary reaction schemes in CFD chemistry submodels. Another key difference is that the furnace is subdivided into regions that reveal the chemical structures of pc flames in detail, as for flames of gaseous fuels. The combustibles mass fraction used to delineate the regions near fuel injectors explicitly accounts for the impact of soot and char combustibles on the burning rates of gaseous fuel compounds and, especially, on NO production rates. These various fuel components are consumed in order of their chemical reaction kinetics, rather than with a priori assumptions such as the mixing-limited volatiles burning rates in CFD. Indeed, the competition for O2 among the various fuels is an essential characteristic of coal flames at any scale, and ChemNet is the most accurate means developed thus far to depict it.

292

Process Chemistry of Coal Utilization

But ChemNet provides much more than a collection of truly comprehensive reaction mechanisms. These analyses also incorporate O2 entrainment rates from the CFD flowfields to represent mixing among regions, as well as mean thermal histories for gases and the effective radiation temperature along each region. Despite the omission of all turbulence aspects, these operating conditions properly sequence the stages in the chemistry both within and among the different regions. And this sequencing is equal in importance to the fidelity of the reaction mechanisms, as seen most clearly in the flame structures that minimize NO production. Regardless of the scale of a furnace, NO forms as soon as the coal reaches a temperature that sustains devolatilization and ignites volatiles, simply because the primary air streams that carry the coal are excessively oxidizing. The early NO accumulates as long as O2 in the primary stream is available to burn out gaseous fuels, char, and/or soot. Once primary O2 falls below a threshold, HCN and NH3 accumulate while NO is reduced into N2 by CO and, perhaps, H2. GHCs ignite and stabilize flames but are superfluous to NO production. Given sufficient time under reducing conditions and a favorable thermal history, near-burner NO can be driven to a minimum near zero and all fixed N-species can form N2. However, this time is determined by O2 entrainment from secondary air streams, which is clearly the key operating factor. Hence, both temperatures and O2 entrainment rates provide the means to control NO emissions, albeit only in part. In practice, the fuel quality impacts are very strong and any rational control strategy must contend with them. The clearest indications of how O2 entrainment rates affect near-burner NO are the huge variations for the ribbons from different coal injectors in the T-fired furnace simulation (cf. Fig. 6.48). The variations in ribbon effluent NO levels are large and those for fixed-N species are even larger. As all injectors had the same coal feedrate, these variations are entirely due to differences among the temperature histories and, especially, the O2 entrainment rates for different ribbons. Rates that are too strong to convert all fixed-N into N2 generate more NO in the ribbon effluents, whereas weak entrainment rates move fixed-N into the quench layer, where these species are converted under oxidizing conditions that produce NO. The minimization of near-burner NO requires a delicate balance between early NO reduction and a later conversion of fixed-N into N2 that can only be struck via tightly regulated O2 entrainment. According to the ChemNet simulations, this target was struck in some, but not all, of the ribbons in the T-fired furnace and also in the pilot-scale CRF flame of PR bituminous coal. Indeed, the flame structures for fuel cores and ribbons in the commercial furnace and the flame core and mixing layer in the pilot-furnace are remarkably similar (as Fig. 6.47 vs. Figs. 6.19 and 6.20). The only noticeable difference is the substantial extent of char burnout in the flame core of the CRF flame, which accelerates the decay in the O2 level toward zero that sets the stage for rapid NO reduction and N2 production. ChemNet simulations of the CRF furnace set the current benchmark on accurate NOX predictions for an extremely broad range of fuel quality. The SSE on predicted NOX for staged CRF flames is 32.4 ppm, which is unbelievably small in the realm of CFD. ChemNet accurately describes how cofiring with biomass that has both much less and much more fuel-N than the coal component will reduce NOX emissions. In contrast, the rudimentary NOX production submodels in CFD cannot predict less NOX when any

ChemNet furnace applications

293

blend component contains more nitrogen than a baseline coal. This performance is especially significant because none of the fuel quality parameters in the simulations were adjusted once their values were specified for the baseline, coal-only cases. Only the proximate and ultimate analyses of the fuels, grind size fractions, and biomass loadings were changed (as in the testing program) to achieve this agreement. The predicted impacts of biomass loading, furnace stoichiometry, and, to a lesser extent, staging are also accurate. It will be interesting to see how long it takes for different simulation strategies to replicate Chemnet’s performance on NOX emissions for such a broad range of fuel quality. ChemNet is especially well-suited for NOX control, VOC emissions, and other aspects of furnace chemistry, such as SO3 levels along gas cleaning systems and the transformations of several trace elements. It resolves char burnout across the PSD with CBK/E, so the method should also be suitable for LOI emissions control. However, as yet, predicted LOI emissions from ChemNet have not been thoroughly validated with test data even while the extensive CRF database on LOI remains to be quantitatively interpreted. The ChemNet applications completed to date expose two drawbacks, one conceptual and one tangible. As the equivalent reactor networks are developed directly from the CFD flow and temperature fields, ChemNet simulations are susceptible to defects in the CFD results. The defects are not confined to small aspects, particularly because the physics that CFD can accommodate are severely constrained by the convergence issues of 3D furnace simulations (cf. the Introduction to Chapter 4). Notwithstanding, there are no alternatives with better accuracy for large coal-fired furnaces, and ChemNet represents a marked improvement over the rudimentary chemistry in CFD chemistry submodels. None of the ChemNet simulations completed to date have flaws that could be traced to problems with the CFD results. ChemNet simulations for large furnaces are inevitably labor-intensive and, sometimes, excessively so. The cases completed thus far constitute the basis to automate most, if not all of the CFD analysis for the reactor network, although the automation is, in itself, a very large undertaking. Fortunately, as demonstrated for the CRF pilotflame, only a handful of CFD simulations were analyzed in detail to cover tests over a much broader operating domain with robust extrapolation procedures. And once bulk flow patterns and regions have been mapped for a particular furnace, the CFD postprocessing for different coals and operating conditions can be streamlined. Nonetheless, ChemNet simulations should only be developed from a clear understanding of why an analysis based on comprehensive chemistry is warranted and which specific information that analysis will deliver. The ChemNet simulations for CFBCs did not validate NOX predictions, because guidance is required from CFD simulations on mixing rates of all streams in a splash zone and on secondary air entrainment rates along risers. However, they did identify a unifying phenomenon for LOI emissions as the potential for ignition to a near-diffusion-limited burning rate across the char PSD. LOI levels increase whenever the ignition is suppressed in any portion of the char PSD and when fully ignited states are extinguished as char moves through the riser. Ignition is suppressed by low intrinsic oxidation reactivities—as with chars from low volatility coals— starting from the smallest char sizes. Fully ignited particles may be extinguished

294

Process Chemistry of Coal Utilization

by thermal annealing with any coal and by ash encapsulation with coals that have excessive ash levels. Thermal soaking at 850°C for a few seconds diminishes char reactivity by nearly two orders of magnitude due to thermal annealing, and the annealing rate surges after ignition due to the much hotter char particle temperatures. But thermal annealing does not seem to undermine the potential to manage LOI by recirculating the largest unconverted char particles through multiple passes through the CFBC because the annealing occurs upstream of the riser even on the first pass.

References Altenkirch RA, Peck RE, Chen SL. Disappearance of nitric oxide and cyanide in onedimensional coal dust/oxidizer flames. Combust Sci Technol 1979;20:49–58. Chen JC, Niksa S. A radiant flow reactor for high-temperature reactivity studies of pulverized solids. Rev Sci Instrum 1992;63:2073–83. Felix LG, Bush PV, Boylan DM, Niksa S, Liu G-S. Development of a validated model for use in minimizing NOX emissions and maximizing carbon utilization when cofiring biomass with coal. In: EPRI-DOE-EPA-A&WMA combined utility air pollution control symposium: the MEGA symp. EPRI, paper no. 21; 2003. Glarborg P, Marshall P. Mechanism and modeling of the formation of gaseous alkali sulfates. Combust Flame 2005;141:22–39. Krishnakumar B, Niksa S. Predicting SO3 levels along utility gas cleaning systems. In: EPRI-DOE-EPA-A&WMA combined utility air pollution control symposium: the MEGA symp., Baltimore, MD, paper no. 16; 2010. Krishnakumar B, Niksa S, Fujiwara N. Predicting selenium emissions from utility gas cleaning systems. In: International conference on air quality IX, UND EERC, Arlington, VA, October; 2013. Liu G-S, Niksa S. Pulverized coal flame structures at elevated pressures. Part 1. Detailed operating conditions. Fuel 2005;84(12/13):1563–74. Liu G-S, Niksa S. A global NOX submodel for pulverized coal flames at elevated pressures. Combust Sci Technol 2006;178(5):953–74. Monroe LS, Clarkson RJ, Stallings J. Comparison of pilot-scale furnace experiments to fullscale boiler performance of compliance coals. In: Book 3, EPRI/EPA 1995 joint symposium on stationary combustion NOX control, Kansas City, MO, May 16–19; 1995. Niksa S. FLASHCHAIN® theory for rapid coal devolatilization kinetics. 8. Modeling the release of sulfur species from various coals. Energy Fuel 2017;31:4925–38. Niksa S. Predicting nitrogen release during coal tar decomposition. Proc Combust Inst 2018;37:2765–72. Niksa S. Simulating volatiles conversion in dense burning coal suspensions. Part 1. Validation of reaction mechanisms. Fuel 2019a;252:821–31. Niksa S. Simulating volatiles conversion in dense burning coal suspensions. Part 2. Extrapolations to commercial p f firing conditions. Fuel 2019c;252:832–40. Niksa S, Cho S. Conversion of fuel-nitrogen in the primary zones of pulverized coal flames. Energy Fuel 1996;10:463–73. Niksa S, Krishnakumar B. Predicting Hg emissions rates with device-level models and reaction mechanisms. In: Granite E, Pennline HW, Senior CL, editors. Mercury emissions control for coal-derived gas streams. Wiley; 2012 [Chapter 27].

ChemNet furnace applications

295

Niksa S, Liu G-S. Incorporating detailed reaction mechanisms into simulations of coal-nitrogen conversion in p. f. flames. Fuel 2002;81:2371–85. Niksa S, Liu G-S. Pulverized coal flame structures at elevated pressures. Part 2. Interpreting NOX production with detailed reaction mechanisms. Fuel 2005;84(12/13):1575–85. Niksa S, Liu G-S, Hurt RH. Coal conversion submodels for design applications at elevated pressures. Part I. Devolatilization and char oxidation. Prog Energy Combust Sci 2003;29:425–77. Niksa S, Padak B, Krishnakumar B, Naik CV. Process chemistry of Br addition to utility flue gas for Hg emissions control. Energy Fuel 2010;24(2):1020–9. Niksa S, Sakurai Y, Fujiwara N. Simulating limestone utilization and SO2 emissions during CFBC. In: Proceedings of 2015 international conference on coal science and technology, IEA, Melbourne, Australia; 2015. Niksa S, Sakurai Y, Fujiwara N. Predicting the conversion efficiencies of any coal type in CFBCs. Energy Fuel 2017;31:4507–19. Peck RE, Midkiff KC, Altenkirch RA. The evolution of nitrogen from pulverized subbituminous coal burnt in a one-dimensional flame. Proc Combust Inst 1984;20:1373–80. Peck RE, Glarborg P, Johnsson JE. Kinetic modeling of fuel-nitrogen conversion in onedimensional, pulverized coal flames. Combust Sci Technol 1991;76:81–109. Schlichting H. Boundary layer theory. 7th ed. New York: McGraw-Hill Book Co.; 1979. Skjoth-Rasmussen MS, Glarborg P, Ostberg M, Johannessen JT, Livbjerg H, Jensen AD, Christensen TS. Formation of polycyclic aromatic hydrocarbons and soot in fuel-rich oxidation of methane in a laminar flow reactor. Combust Flame 2004;136:91–128. Uijttewaal WSJ, Oliemans RVA. Particle dispersion and deposition in direct numerical and large eddy simulations of vertical pipe flows. Phys Fluids 1996;8(10):2590–604. Yang H, Yue G, Xiao X, Lu J, Liu Q. 1D modeling on the material balance in a CFBC boiler. Chem Eng Sci 2005;60:5603–11. Yilmaz A, Hindiyarti L, Jensen AD, Glarborg P, Marshall P. Thermal dissociation of SO3 at 1000–1400 K. J Phys Chem A 2006;110(21):6654–9. Zhao X, Lu J, Yang J, Zhang Q, Dong F, Yu L, Yang Z, Yue G. Operational performance and optimization of a 465t/h CFB boiler in China. In: Proc. 18th international conference on fluidized bed combust. New York: ASME; 2005. p. 791.

ChemNet gasifier applications

7

Nomenclature CC %CAR FE %Hdry LC %OAR %Odry t

total flowrate of carbon in coal, kg/h C-content of coal on an as-received basis fraction of a stream entrained into the process flow to time t H-content of coal on a dry basis fractional suspension loading, as a ratio of flowrates of coal to gaseous feedstreams O-content of coal on an as-received basis O-content of coal on a dry basis time, s

Greek symbols βM ηC ηM τM

empirical mixing constant for entrainment conversion efficiency of coal conversion efficiency of CH4 time lag for entrainment

This chapter presents ChemNet case studies that characterize the chemical structures and syngas compositions for entrained-flow gasifiers. It contains one case each for lab-, pilot-, demonstration-, and commercial-scale systems. The goal is to demonstrate that ChemNet delivers predictions for conversion efficiencies and syngas compositions, including minor species like CH4, that are appreciably more accurate than CFD results. At face value, this goal can be met in straightforward comparisons between measured and predicted values for the output variables of interest. The next section discusses the multitude of factors that can undermine “straightforward comparisons” with data from any gasifier at any scale. Then case studies are presented in sequence for gasifier configurations at a progressively larger scale.

7.1

ChemNet validations with data from gasifiers

The previous chapter began in Section 6.1 with a survey of measurement uncertainties for pc furnaces at various scales, including ambiguous thermal histories for the coal suspension, asymmetries in the flow fields, and very brief conversion time scales in NBFZs. The ultimate conclusion is that the potential to validate simulation results diminishes and, ultimately, vanishes in the progression from the lab- to pilot- to commercial-scale furnaces. The measurement uncertainties are even larger in entrained-flow gasifiers for several reasons. Gasifiers operate at pressures as high Process Chemistry of Coal Utilization. https://doi.org/10.1016/B978-0-323-89959-8.00008-2 Copyright © 2022 Elsevier Ltd. All rights reserved.

298

Process Chemistry of Coal Utilization

as 10 MPa and temperatures to 2500°C, and vessel walls are often used to convey molten slag to collection vessels. Extractive sampling through probes is impossible in such hostile environments. Even if it was feasible, the results would defy quantitative interpretations because the flowfields are so complex. In pc furnaces, burners and injectors are designed to regulate mixing rates among the primary coal suspensions and secondary and tertiary air streams. As seen in the previous chapter, the resulting flow structures across an NBFZ are three-dimensional and complex, but nonetheless amenable to reasonably detailed characterization with CFD. In contrast, the nearinjector zones in entrained-flow gasifiers are designed to ignite the coal and burn away oxygen as fast as possible by imposing the fastest mixing rates, to achieve the hottest temperatures. This is often accomplished with injectors for oxidizer and steam that are off-radius, impinging coal injectors, and intense recirculation of hot syngas from downstream into the coal jets. The resulting flow structures from CFD are anything but coherent and subject to gross distortions from minor perturbations on the inlet conditions. As a tangible illustration, consider the ranges of conditions near the fuel injectors reported in the literature for two of the most popular gasifier designs by Shell and General Electric Power Systems (GEPS). In Shell gasifiers, coal in very fine grinds is entrained in N2 or CO2 at 1.0 kg-coal/kg-gas and co-fed with O2 through four nozzles in an injection plane near the base of the reactor. The nozzles are offset from the reactor radius so that the horizontal coal jets impinge into a rectangle that rotates the flow upward in a helical pattern (Cao et al., 2018), somewhat like the flowfield in T-fired pc furnaces. In GEPS gasifiers, the coal-water slurry is fed at 2.0 kg-coal/kg-water downward through a cylindrical annulus into the reactor vessel. The coal jet is cofed with a swirled O2 stream that mixes into the coal jet after the water evaporates. In both designs, once the coal suspension leaves the injector nozzle, it heats primarily by radiation from the coal flame and entrainment of hot syngas, and from the heat release from fuels expelled from the coal. Among five recent CFD simulation studies of Shell gasifiers, three invoked instantaneous combustion within the fuel injection plane (Sun et al., 2011; Lee et al., 2014; Zhou et al., 2018), and the other two gave grossly different results (Cao et al., 2018; Gazzani et al., 2013). Gazzani et al. (2013) reported coal injection velocities approaching 200 m/s and extremely rapid coal heating at rates exceeding 105°C/s. Transit times from the injector nozzles to the upward helical flow were only a few milliseconds. In contrast, the coal velocities reported by Cao et al. (2018) never exceed 10 m/s, so that transit times within the horizontal injection plane were a few hundred milliseconds, which was sufficient to consume most of the O2. Only one set of these characteristics can be realistic but which one is it? Despite the simpler flow structures in GEPS gasifiers, the operating conditions are even more uncertain than those for Shell gasifiers. As the O2 co-flow is swirled, the primary coal jet can be resolved into an internal recirculation zone (IRZ) immediately downstream of the nozzle, surrounded by the expanding coal jet which, in turn, is surrounded by an ERZ for hot syngas (Monaghan and Ghoniem, 2012). Ultimately, these regions coalesce into plug flow into the reactor outlet. The main source of ambiguity is that the swirl intensity determines the strengths of the recirculation zones which, in

ChemNet gasifier applications

299

turn, govern the entrainment of hot syngas into the coal suspension. A recent parametric study on the swirl number gave combustion zones that extended only 0.5 m from the nozzle for strong swirl to nearly the entire reactor for weak swirl (Bi et al., 2015). Moreover, the only simulation study that resolved the heat-up of the coal jet gave nearinstantaneous heating (Monaghan and Ghoniem, 2012), which is implausible for such heavy coal loadings. The recycle ratio from the ERZ into the entire coal jet was two, but the authors did not resolve the recycle portion into the IRZ, where most O2 is consumed. Such ambiguities in the flow structures propagate into the assigned thermal histories for the coal suspensions which, in turn, affect the predicted coal conversion before O2 is eliminated. In the midst of such enormous ambiguities, there is little incentive to even contemplate quantitative validations of gasifier simulation results with any data other than outlet syngas compositions for broad ranges of coal quality. However, syngas compositions are integral in nature and can, therefore, be matched by an infinite number of trajectories through the excluded, upstream portion of the system. In other words, an infinite number of mathematical models can accurately match characteristics at the exit of a system even though their dynamics are wildly different. Nonetheless, one still can assert that agreement among measured and predicted exit characteristics are within measurement uncertainties, provided that the suite of measurements close the mass and element balances. But this agreement does not establish the validity of the reaction mechanisms in the simulation; instead, it only establishes that the mechanisms are viable candidates, pending more stringent validation, presumably, at a lab-scale where the dynamics may be more accessible. Conversely, discrepancies with the data mean that the simulations did not sustain the evaluation, and the reaction mechanisms are suspect. It is probably best to regard detailed flow patterns and conversion histories as presentation cartoons with little, if any, quantitative validity. To reiterate from Section 6.1, this author describes agreement among data and simulations under these circumstances with the phrase, “within useful quantitative tolerances.” Presumably, simulations were developed in the first place to demonstrate that they can reproduce the measured syngas compositions within some externally set tolerance, where the tolerances are simple metrics without any qualifications at all to emphasize the practical utility over and above the statistical variance. This is precisely the distinction drawn by the “useful” in “within useful quantitative tolerances.” Gasification reaction mechanisms should be validated in simplified flowfields at the lab-scale to demonstrate that they are able to accurately depict the reaction dynamics, preferably over the full domain of conditions in the commercial application of interest. Reactor network simulations are the only means currently available to do this with truly comprehensive reaction mechanisms. Once the validation is sustained at lab-scale, ChemNet simulations with the same mechanisms can be validated with datasets from pilot-scale to contend with some of the complications that arise at commercial-scale, especially turbulent mixing effects. In the ideal situation, a CFD simulation of the pilot-scale flowfield carries all the essential aspects of mixing and entrainment into the equivalent reactor network in the ChemNet simulation. When the same coals are tested at lab- and pilot-scale, there may be no adjustable parameters whatsoever in the validations with datasets from pilot-scale. Otherwise, there can be

300

Process Chemistry of Coal Utilization

up to two adjustable parameters for conventional gasifier performance indices, which are the initial reactivities for char conversion by oxidation and gasification (because char-N conversion into NO is superfluous in gasifiers). In either situation, agreement with datasets from pilot-scale can demonstrate that the simulations accurately predict characteristics at the system exit within useful quantitative tolerances, but it does not validate the mechanisms, per se. Once a set of mechanisms has performed as well as necessary at a pilot-scale, it can be used for commercial-scale systems. But comparisons among measurements and simulation results have nothing to do with validations of reaction mechanisms or any other aspects of the simulations. For reasons already enumerated, any agreement between field test data and simulation results does not firmly establish anything about the simulations or the reaction mechanisms. When all is said and done, ChemNet simulations of commercial gasifiers give an educated idea of when and where coal is converted and minor syngas components form; by which reaction channels these processes occur; and how the conditions can be regulated to improve the overall performance. But they really have very little to do with quantitatively validating any aspect of the simulations at a commercial-scale.

7.2

Target variables for ChemNet simulations

Specifically, what are ChemNet simulations of gasifiers good for? That is, what species concentrations, conversion indices, etc. can be accurately determined with a ChemNet gasifier simulation? This section addresses this question in terms of the coal conversion, minor syngas components, and heteroatomic speciation. The reactants for coal gasification largely comprise coals’ moisture, carbon, oxygen, and hydrogen. However, the reactants exclude mineral matter and all minor and trace elements. In gasifiers, the major elements form mixtures of CO, CO2, H2, and steam called “syngas.” For the very hot temperatures in entrained-flow gasifiers, syngas compositions at the quench point can be accurately estimated from a coal composition, all stream flowrates, and a thermochemical equilibrium gas composition at some threshold temperature along the gas quench cycle provided that nearly all coal is converted. Consequently, resolving chemical reaction mechanisms is superfluous to the concentrations of all four major components in syngas. Unless the UBC is truly excessive, the flue gas composition is hardly perturbed by unconverted coal. Notwithstanding, UBC emissions from gasifiers are legitimate targets for ChemNet simulations. All the mechanisms for drying, devolatilization, tar decomposition, homogeneous reforming, and conversion of char and soot via both oxidation and gasification are needed to predict UBC emissions. Only the homogeneous mechanism may be streamlined by eliminating the reactions for N-species transformations. Indeed, CBK/E and CBK/G are among only a handful of char conversion mechanisms that describe the various deactivation modes during the latter stages of char conversion, which is essential to accurate UBC predictions at the very hot temperatures in

ChemNet gasifier applications

301

gasifiers. Unfortunately, UBC predictions also require accurate thermal and concentration histories for all particle trajectories through a reactor, which is definitely challenging. In syngas, CH4 and GHC levels often make disproportionate contributions to the cold gas efficiencies of gasifiers operated at moderate temperatures. As seen later in this chapter, ChemNet simulations delivered the first and only reliable predictions for CH4 concentrations in syngas at both pilot- and commercial-scale. The transformations of coal-S into H2S, COS, and CS2 during gasification are well-suited for ChemNet simulations, albeit with qualifications. The greatest ambiguities in the chemistry pertain to the release of coal-S during devolatilization. In FLASHCHAIN, organic-S must be allocated into aliphatic and aromatic S-functional groups even though the supporting analytical methods to do this are not widely available (Niksa, 2017). Consequently, the devolatilization of coal-S cannot currently be accurately predicted from a coal’s proximate and ultimate analyses; specialized analytical support is required. This is not a major hindrance in gasifier applications with very high H2 pressures, under which essentially all coal-S will ultimately be present in the nascent syngas. Otherwise, the partitioning of coal-S among gases, char, and mineral forms requires analytical support for every coal sample. Fortunately, validated homogeneous reaction mechanisms for S-transformations under reducing atmospheres are available. In summary, ChemNet postprocessing has already been used to accurately predict CH4 levels in syngas, as seen below. It also gives reasonable estimates for coal conversion and UBC emissions, subject to two qualifications. First, the initial oxidation and gasification reactivities must be specified in calibration for every coal. Second, UBC emissions can only be predicted when the primary coal trajectories through a reactor are accurately characterized, which is usually impossible in commercial reactors. The mechanisms to predict S-speciation in syngas are already available. However, the concentrations of S-species in syngas remain to be validated with measured values because specialized analytical support is required to specify the distributions of the major S-forms in every coal sample. There is little incentive to pursue other minor and trace species in syngas because the excessively reducing conditions usually skew their speciations toward only one or, at most, two species. Such is the case with the nitrogen, halogens, alkali/alkaline earth metals (AAEMs), and Hg in coal.

7.3

Case studies on entrained-flow coal gasification

This section presents a series of case studies on coal conversion during entrained-flow coal gasification. It moves through cases at lab-, pilot-, demonstration-, and commercial-scale. However, only the lab-scale case formally specified the equivalent reactor network from a conventional CFD simulation; otherwise, the ChemNet simulations are based on simplified, approximate networks. The goal is to resolve the chemical structures of various gasification environments and to demonstrate accurate predictions for coal conversion and syngas composition, including CH4.

302

Process Chemistry of Coal Utilization

7.3.1 CH4/coal co-gasification at lab-scale In the coal-and-gas-to-liquids (CGTL) process, coal and natural gas are simultaneously converted into syngas, which is cleaned and converted to methanol, which is then used to produce transportation fuels (Lim et al., 2013). The advantage of the CGTL scheme over conventional Fisher-Tropsch Synthesis (FTS) is that it uses natural gas as the source of hydrogen, so that no CO2 is emitted. Moreover, the yield of liquid product from solid-phase carbon is at least four times that from FTS, because for every coal carbon, this process also converts one carbon from gaseous methane into liquid. This process intensification is key to the vastly reduced capital requirement. The use of natural gas instead of water as the primary source of hydrogen also reduces the overall energy requirement. By simultaneously steam-reforming natural gas and gasifying coal, the product syngas has a CO:H2 ratio of 1:2, as required for methanol production. The water into the gasifier is regenerated in the methanol-toolefin step and recycled.

7.3.1.1 Interpretations of lab-scale gasifier data A process design and economic analysis were developed for CGTL from lab-scale testing on CH4/coal cogasification with a single subbituminous coal. Conversions of both coal and CH4 were monitored over a broad domain of inlet gas compositions and temperature at 3.0 MPa. The test system was an expanded version of the p-RCFR described in Section 6.3.1.2 in connection with coal flames at elevated pressure. For the gasification tests, the radiant igniter in the p-RCFR was connected to three modular heated sections to extend residence times to almost 1.5 s in three increments. Coal loadings were varied from 3.1 to 9 wt.%. The carrier gas contained from 9% to 28% steam; from 2.5% to 8% CH4; and from 20% to 30% H2; with a balance of Ar. These H2 levels were sufficient to produce negligible levels of soot in the solid products when expressed as a fraction of the total carbon fed into the reactor. Temperatures in the modular heaters were varied from 1400°C to 1500°C, while the igniter was kept at 1500°C. The primary variables were the coal loadings and steam and CH4 levels, with smaller variations in temperature and H2 level. ChemNet simulations were developed, first, to quantitatively interpret the measured conversion across the test domain and, second, to extrapolate to commercial gasifier conditions to provide input for the design and economic analyses. Thermal histories and transit times for the tests were specified from CFD simulations. As O2 was excluded from the entrainment stream, none of the steep gradients and turbulent transport in the flows for pressurized coal combustion arose in these tests. The flowfield was primarily in turbulent plug flow downstream of the igniter section. The coal was heated at 3500°C/s in the igniter, where the volatiles were hydrogenated by H2 and reformed in steam to produce the gasification agents that converted char along the modular heaters. Table 7.1 compares the distributions of primary volatiles, secondary pyrolysis products, and secondary volatiles hydrogenation products from a subbituminous coal. Primary devolatilization converts about half the coal into char.

ChemNet gasifier applications

303

Table 7.1 Predicted volatile product distributions from subbituminous in daf wt.%.

Volatiles Tar Soot PAH Oils Char H2 O CO2 CO H2 CH4 C2H6 C2H4 C2H2 C3H6 HCN H2 S

Primary

20 Pyrolysis

20 Hydrogenation

51.6 9.8 – – – 48.4 3.6 10.4 14.6 2.5 4.5 0.7 1.9 – 1.6 1.0 0.6

51.6 – 8.8 – – 48.4 3.6 10.4 16.3 4.2 0.6 – – 5.6 – 1.1 0.6

51.6 – – (7.2) 0.7 48.4 3.8 10.8 15.4 1.2 7.3 0.8 6.6 – 1.7 1.3 0.6

Primary products are the products that leave the coal phase before they undergo any secondary conversion. These products contain tar albeit at a relatively low yield of about 10 daf wt.% because of the elevated pressure and high O-content of this particular coal. The latter feature explains why this distribution is dominated by oxygenated gases, which comprise about 60% of the ultimate volatiles yield. Minor amounts of GHCs and heteroatom species round out the distribution. The secondary volatiles pyrolysis products have no tar but do contain a similar amount of soot plus the CO, H2, and HCN released during tar decomposition. The total yield of GHCs is diminished due to their addition to the soot phase to nearly compensate for CO elimination, and the GHC distribution has been reduced to a C2H2/CH4 mixture with an abundance of C2H2 for this particular coal. The hydrogenation products contain neither tar nor soot because elevated H2 pressures hydrogenate primary tars into GHCs and oils, as described elsewhere (Niksa, 2018a). The PAH yield in Table 7.1 is the primary tar yield adjusted for the spontaneous release of tar-O, tar-H, and tar-N. It appears in parentheses to denote that PAH is ultimately hydrogenated into oils, C2H4, and CH4. PAH hydrogenation is represented as the global process of Eq. (4.10), based on reinterpretation of the data reported by Nelson and Huttinger (1986), which gives 55% CH4, 35% C2H4, and 10% oils. Consequently, this distribution contains no product heavier than oils. Its levels of oxygenated gases are comparable to the others because all tar-O is hydroconverted into CO, but it has the greatest levels of GHCs, by far.

304

Process Chemistry of Coal Utilization

In the ChemNet simulations, the volatiles were instantaneously converted into secondary hydrogenation products, because of two observations in the tests. First, the yields of soot were always negligible fractions of the total carbon fed into the reactor. Second, the contribution to the carbon inventory in the gas phase from CH4 was always much greater than that from tar due to the low suspension loadings. Moreover, the extremely high H2 partial pressures in these tests—typically greater than 0.9 MPa—promote rapid and complete hydrogenation, particularly at elevated test temperatures (Niksa, 2018a). The reforming chemistry in the gas phase is described with the mechanism described in Section 5.4.1 with 566 reactions among 154 species, including all GHCs up to benzene, toluene, and phenol. Char gasification by steam with inhibition by CO and H2 is described with CBK/G. In addition, char hydrogasification by H2 is described in the CBK framework with a single, nth-order reaction that acts in parallel with steam gasification. As explained below, it is not negligible under these test conditions. The ChemNet equations for reaction systems with steam and H2 were presented in Section 5.5.3. Preliminary ChemNet simulations that omitted char conversion clarified several aspects of the reforming chemistry. Virtually no reforming chemistry occurs until the stream reaches about 1250°C as it leaves the igniter stage. Thereafter, the gas compositions change continuously in proportion to the transit time. There are no abrupt surges in any of the compositions while CH4 is reformed by steam into H2 and CO with trace amounts of CO2. A transit time of 1.4 s is sufficient to equilibrate the syngas composition, provided that the temperature is at least 1500°C, as seen in Table 7.2. All three cases had identical inlet gas compositions. The syngas composition is not equilibrated at 1400°C but more closely approaches the equilibrium composition at 1450°C and nearly achieves it at 1500°C. Consequently, syngas from the bulk of the tests for 1400°C will contain appreciable levels of intermediates and major products in proportions that differ significantly from the equilibrium products. Preliminary simulations also found that CH4 conversion grew from 20% to 50% while the steam level was increased from 9% to 28%, and from 60% to 95% when the temperature was increased from 1400°C to 1500°C. Table 7.2 Comparison of equilibrium and kinetically limited product compositions at three reactor temperatures. 1400°C

CH4 H2O H2 CO CO2

1450°C

1500°C

Equil.

Reform. chem.

Equil.

Reform. chem.

Equil.

Reform. chem.

59 ppm 0.118 0.489 0.090 0.006

0.035 0.165 0.414 0.058 0.007

11 ppm 0.161 0.380 0.072 0.009

0.011 0.176 0.354 0.062 0.009

7 ppm 0.161 0.379 0.072 0.008

0.003 0.164 0.373 0.070 0.008

ChemNet gasifier applications

305

The first attempt to interpret the reported conversions combined the reforming chemistry with char gasification by steam only. An initial gasification reactivity was specified to accurately depict the extents of char conversion across the test domain. However, the predicted CH4 conversions were uniformly greater by 30% than the measured conversions across all inlet gas compositions and all test temperatures. These results also overestimated the sensitivity of CH4 conversion to the steam level. Most of the discrepancies were eliminated by simulating coal conversion via devolatilization plus char gasification by both steam and H2, with continuous reforming chemistry in the gas phase among all products. In principle, char hydrogasification rectifies overpredicted CH4 conversions because it directly produces CH4, especially since the H2 concentrations in the tests increase continuously, and there is little time for the CH4 from hydrogasification to be eliminated by homogeneous reforming. In contrast, there has been widespread agreement in the coal research community that char hydrogasification is orders of magnitude slower than steam gasification, particularly at the elevated temperatures in this application. In actuality, char hydrogasification is not negligible in comparison with steam gasification, provided that the H2 partial pressure approaches 1 MPa (Niksa, 2018b). Most of the tests under consideration satisfy this stipulation. The ultimate interpretation of this database used the default char gasification reactivity parameters for both steam and H2 for the subject coal, without further adjustment. These values give char conversion mostly from steam gasification with char conversions from hydrogasification of 15% to 18%. Predicted extents of conversion for coal and CH4 are evaluated in Fig. 7.1. The predicted coal conversions are accurate across the entire range of test conditions, in so far as the parity line in the figure is nearly the same as the regression line through the plotted points. The correlation coefficient exceeds 0.8 and the std. dev. is 3.4%, which is probably comparable to the

100

90

80 60 40

80 20

Runs 77 – 94 SG Series r2 = 0.60 s = 23.5

70

60 60

70

80

90

MeasuredXCOAL,%

100 –40

–20

0

20

40

60

80

0 –20

–40 100

MeasuredXCH ,% 4

Fig. 7.1 Evaluation of the predicted extents of (left) coal and (right) CH4 conversion via devolatilization plus char gasification by steam and H2 based on default char reactivity parameters.

4

Runs 77 – 94 SG Series r2 = 0.81 s = 3.4%

XCH ,% W/H2O & H2 Gasification

Predicted XCOAL,%

100

306

Process Chemistry of Coal Utilization

measurement uncertainties. However, the predicted CH4 conversions are not particularly accurate. They overpredict the lowest and greatest measured conversions but have reasonably accurate predictions for intermediate CH4 conversions. The std. dev. is greater than 20%.

7.3.1.2 Extrapolations to commercial processing There are three important differences among the operating conditions evaluated in the lab-scale tests and commercial processing conditions: (1) Coal suspension loadings are much greater; (2) The reactor pressures are much higher at 5 to 7 MPa; and (3) The feedstream cannot be diluted with inert gases. In addition, the residence times through the gasifier will probably be longer than the 1.4 s imposed in the lab tests, because commercial entrained coal gasifiers typically operate with residence times of 2.5 to 4 s. These issues were clarified with the same idealized reactor network in the previous section and the same mechanisms and parameters. Whereas the reaction mechanisms are realistic, the descriptions of mixing and fuel particle dispersion in the analysis are highly idealized for any application at a commercial scale. The coal suspension loading in commercial applications is not an independent operating condition because the economic performance is based on feeding 60% of the process carbon as coal and the remainder as natural gas. This stipulation determines the suspension loading as follows: 2

3 CC 6 7 %CAR ηC  7 Lc ¼ 286 48CC 8CC
18 8CC %CAR %OAR 5  + + 9ηM 9ηM 16 %CAR ηC 12 16

(7.1)

where LC is the suspension loading, as a ratio of flowrates of coal to gaseous feedstreams; CC is the total flowrate of carbon in coal; %CAR and %OAR are the C- and O-contents of coal on an as-received basis; and ηC and ηM are the conversion efficiencies of coal and natural gas. For the subbituminous coal, the loading is 46.2 wt. % when only CH4 and the stoichiometric amount of steam are used as the entrainment fluid, and both conversion efficiencies are 90%. This loading represents the theoretical maximum, because all commercial conditions have excess steam and additional H2 in the feed to control sooting during CH4 reforming. For the case with the stoichiometric level of steam and no added H2, the steam and natural gas flowrates are 1.55 and 0.62 kg/s, respectively. Three simulation cases used 15% excess steam with H2 added at one, two, and three times the CH4 level. A fourth case used an external reformer with H2 added at twice the CH4 level, in which the reformer converted 70% of the CH4 upstream of the gasifier. All cases were run at 1500°C under 3 MPa for a residence time of 1.4 s, as in the lab-scale tests. The predicted CH4 conversions diminished for progressively more H2 in the feed, whereas the coal conversions increased but not by enough to compensate. Compared to the most similar conditions in the lab-scale database, the CH4 conversions are lower by up to 10%, and the coal conversions are lower by up to 9%. The contribution of

ChemNet gasifier applications

307

hydrogasification to char conversion is more-than-double, at 40%. Reforming natural gas upstream of the gasifier converts essentially all the CH4 feed but lowers the coal conversion by 1.5%. Commercial conditions with realistic suspension loadings and no diluents give slightly lower conversions of both natural gas and coal, because relatively rapid CH4 reforming eliminates steam that would otherwise gasify the char. Consequently, hydrogasification accounts for more of the char conversion but occurs at rates slower than for steam gasification. This feature suggests that conversions may be improved by extending residence times and by elevating the pressure (to accelerate char conversion kinetics). Adding an upstream reformer markedly improved CH4 conversions but probably requires even longer residence times in the gasifier to meet a target coal conversion, which increases capital costs. The survey of commercial processing conditions has temperatures of 1300°C, 1400°C, or 1500°C; pressures of 5.5 or 6.9 MPa; excess steam levels of 15% and 30%; and reaction times of 1, 2, and 4 s. All cases have a suspension loading based on Eq. (7.1) and the feedstreams contain no diluents. All CH4 is reformed in the same gasifier as the coal. Hydrogen is added to manage soot production during CH4 reforming and also to satisfy the larger H2 requirement of hydrogasification compared to steam gasification of char. Hydrogen constituted from 13% to 25% of the feedstream and the levels grew for progressively longer reaction times to sustain greater extents of char conversion through longer times. The predicted CH4 conversions for 1500°C vary from just under 90% to 99% and increase for progressively longer reaction times. They are slightly greater at the greater excess steam level but only by 2% to 3%. The impact of raising the pressure is even weaker. Considering that there was more H2 in the feed for longer reaction times, and that CH4 conversions diminish for progressively greater H2 levels in the feed, the impact of longer reaction times is even greater than it appears to be in this series of simulations. Reaction time is also the dominant factor in coal conversions. They are about 68% in 1 s for all pressures and steam levels and reach complete conversion in 2 s. The contribution from hydrogasification grows for progressively longer reaction times and becomes dominant for times longer than 2 s. This shift reflects variations in the gas compositions for longer times, which tend to produce H2 at the expense of H2O; neither the excess steam level nor pressure affected these contributions or the overall coal conversions. At 1500°C, a gasifier with a 2 s residence time at the lower pressure and excess steam level converts over 90% of natural gas and essentially all the coal in the feedstream. For 1400°C, the performance is the same provided that residence times are extended to 4 s. The only difference is that hydrogasification makes a smaller contribution to char conversion. Nothing in these results undermines the potential for further process development. On the contrary, the required gasifier conditions are well within the range of commercial practice. Whereas 5.5 MPa is higher than the 4.1 MPa for GEPS gasifiers, 1500°C is much cooler than the operating temperatures of both Shell and GEPS gasifiers. And even if the residence time is doubled to 4 s to manage mixing aspects, it is still within the range of commercial practice. The 2013 process economics analysis predicted that diesel fuel can be produced for $2.81/gal (Lim et al., 2013). The capital cost for a plant

308

Process Chemistry of Coal Utilization

producing 100,000 barrels/day of diesel fuel was estimated at $3.18 billion or $31,800/daily barrel of capacity, less than half of the estimated $70,000/daily barrel for a coal-to-liquids plant based on FTS. Of course, the economics would need to be re-evaluated whenever the technical performance was further established at a pilotscale.

7.3.2 Coal gasification at pilot-scale This section presents ChemNet simulations for a broad testing program in a pilot-scale entrained-flow gasifier operated by CanmetENERGY (CEN). The test conditions supported the development of the E-Gas technology licensed by CB&I Corp. This gasifier processes coal slurry in a novel two-stage configuration to first generate syngas in an extremely hot, O2-blown first stage, and then to enhance the syngas composition with coal slurry gasification in a second stage at moderate temperatures. Accordingly, the CEN gasifier was operated in a staged mode in which primary syngas was generated with an extremely rich butanol/O2 flame, and coal slurry was injected into the hot syngas within a down-fired, cylindrical reactor to simulate the second stage. The test conditions were selected to reproduce the ultimate syngas from the second stage of the commercial gasifier. One of the primary advantages of this gasification configuration is that the second stage adds GHCs to the syngas which significantly boosts its calorific value. The main objective of the ChemNet simulations is to demonstrate accurate predictions for the syngas composition, including the levels of GHCs.

7.3.2.1 Test conditions The CEN gasification pilot-plant is described elsewhere (Duchesne et al., 2014) albeit in the normal, single-stage operating mode. For the two-stage configuration, the reactor consists of a feeder manifold that supports a butanol/O2 flame; an upper reactor section with injection ports for a steam quench and coal slurry; a lower reactor section with two ports for sample recovery; and a final quench and exhaust section. The reactor ID is 25.4 cm throughout. The butanol/O2 flames were run under extremely fuelrich conditions (0.30