Hydraulic Fracturing and Well Stimulation (Sustainable Energy Engineering) [1 ed.] 1119555698, 9781119555698

Citation preview

Hydraulic Fracturing and Well Stimulation

Scrivener Publishing 100 Cummings Center, Suite 541J Beverly, MA 01915-6106 Publishers at Scrivener Martin Scrivener ([email protected]) Phillip Carmical ([email protected])

Book Series Sustainable Energy Engineering

Hydraulic Fracturing and Well Stimulation

Edited by

Fred Aminzadeh

This edition first published 2019 by John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, USA and Scrivener Publishing LLC, 100 Cummings Center, Suite 541J, Beverly, MA 01915, USA © 2019 Scrivener Publishing LLC For more information about Scrivener publications please visit www.scrivenerpublishing.com. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, except as permitted by law. Advice on how to obtain permission to reuse material from this title is available at http://www.wiley.com/go/permissions. Wiley Global Headquarters 111 River Street, Hoboken, NJ 07030, USA For details of our global editorial offices, customer services, and more information about Wiley products visit us at www.wiley.com. Limit of Liability/Disclaimer of Warranty While the publisher and authors have used their best efforts in preparing this work, they make no representations or warranties with respect to the accuracy or completeness of the contents of this work and specifically disclaim all warranties, including without limitation any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives, written sales materials, or promotional statements for this work. The fact that an organization, website, or product is referred to in this work as a citation and/or potential source of further information does not mean that the publisher and authors endorse the information or services the organization, website, or product may provide or recommendations it may make. This work is sold with the understanding that the publisher is not engaged in rendering professional services. The advice and strategies contained herein may not be suitable for your situation. You should consult with a specialist where appropriate. Neither the publisher nor authors shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages. Further, readers should be aware that websites listed in this work may have changed or disappeared between when this work was written and when it is read. Library of Congress Cataloging-in-Publication Data ISBN 978-1-119-55569-8 Cover image: Oil Rig - Luchschen | Dreamstime.com Cover design by Kris Hackerott Set in size of 11pt and Minion Pro by Manila Typesetting Company, Makati, Philippines Printed in the USA 10 9 8 7 6 5 4 3 2 1

Contents Foreword

xiii

Part 1: Introduction

1

1 Hydraulic Fracturing, An Overview Fred Aminzadeh 1.1 What is Hydraulic Fracturing? 1.2 Why Hydraulic Fracturing is Important 1.3 Fracture Characterization 1.4 Geomechanics of Hydraulic Fracturing 1.5 Environmental Aspects of Hydraulic Fracturing 1.6 Induced Seismicity 1.7 Case Study: Fracturing Induced Seismicity in California 1.8 Assessment of Global Oil and Gas Resources Amenable for Extraction via Hydraulic Fracturing 1.9 Economics of HF 1.10 Conclusions Acknowledgement References

3

Part 2: General Concepts 2 Evolution of Stress Transfer Mechanisms During Mechanical Interaction Between Hydraulic Fractures and Natural Fractures Birendra Jha 2.1 Introduction 2.2 Physical Model 2.3 Mathematical Formulation 2.4 Numerical Model 2.5 Simulation Results

4 5 8 11 14 18 23 27 27 28 30 30

35 37 37 39 40 43 44

v

vi

Contents 2.6 Effect of Hydraulic Fracturing on Natural Fractures 2.7 Conclusion References

3 Primer on Hydraulic Fracturing Concerning Initiatives on Energy Sustainability Michael Holloway and Oliver Rudd 3.1 Hydraulic Fracturing 3.1.1 Environmental Impact – Reality vs. Myth 3.1.2 The Tower of Babel and How it Could be the Cause of Much of the Fracking Debate 3.1.3 Frac Fluids and Composition 3.1.4 Uses and Needs for Frac Fluids 3.1.5 Common Fracturing Additives 3.1.6 Typical Percentages of Commonly Used Additives 3.1.6.1 Proppants 3.1.6.2 Silica Sand 3.1.6.3 Resin Coated Proppant 3.1.6.4 Manufactured Ceramics Proppants 3.2 Additional Types 3.3 Other Most Common Objections to Drilling Operations 3.3.1 Noise 3.4 Changes in Landscape and Beauty of Surroundings 3.5 Increased Traffic 3.6 Chemicals and Products on Locations 3.6.1 Material Safety Data Sheets (MSDS) 3.6.1.1 Contents of an MSDS 3.6.1.2 Product Identification 3.6.1.3 Hazardous Ingredients of Mixtures 3.6.1.4 Physical Data 3.6.1.5 Fire & Explosion Hazard Data 3.6.1.6 Health Hazard Data 3.6.1.7 Reactivity Data 3.6.1.8 Personal Protection Information 3.7 Conclusion Bibliography

46 49 50 53 54 54 55 57 57 58 60 61 63 65 65 66 66 67 68 69 70 72 73 73 74 74 75 76 76 77 77 78

Contents 4

A Graph Theoretic Approach for Spatial Analysis of Induced Fracture Networks Deborah Glosser and Jennifer R. Bauer 4.1 Background and Rationale 4.2 Graph-Based Spatial Analysis 4.2.1 Acquire Geologic Data and Define Regional Bounding Lithology 4.2.2 Details of the Topological Algorithm 4.2.2.1 Data Acquisition, Conditioning and Quanta 4.2.2.2 Details of the k-Nearest Neighbor Algorithm 4.2.3 The Value of the Topological Approach Algorithm 4.3 Real World Applications of the Algorithm 4.3.1 Bradford Field: Contrasting the Graph-Based Approaches; k Sensitivity 4.3.1.1 Data Sources 4.3.1.2 Results 4.3.2 Armstrong PA: Testing the Algorithms Against a Known Leakage Scenario 4.3.2.1 Data Sources 4.3.2.2 Results 4.4 Discussion 4.4.1 Uses for Industry and Regulators 4.5 Conclusions Acknowledgements References

Part 3: Optimum Design Parameters 5 Fracture Spacing Design for Multistage Hydraulic Fracturing Completions for Improved Productivity D. Maity, J. Ciezobka and I. Salehi 5.1 Introduction 5.2 Method 5.2.1 Impact of Natural Fractures 5.2.2 Workflow 5.2.3 Model Fine-Tuning 5.2.4 Need for Artificial Intelligence

vii 79 80 83 84 85 85 86 86 87 87 88 88 88 90 90 91 93 93 94 94

99 101 101 103 104 107 108 109

viii

Contents 5.3 Data 5.4 Results 5.4.1 Applicability Considerations 5.5 Concluding Remarks Acknowledgement References

6

7

Clustering-Based Optimal Perforation Design Using Well Logs Andrei S. Popa, Steve Cassidy and Sinisha Jikich 6.1 Introduction 6.2 Objective and Motivation 6.3 Technology 6.4 Clustering Analysis 6.4.1 C-Means (FCM) Algorithm 6.5 Methodology and Analysis 6.5.1 Available Data 6.6 Applying the FCM Algorithm 6.7 Results and Discussion 6.8 Conclusions Acknowledgements References Horizontal Well Spacing and Hydraulic Fracturing Design Optimization: A Case Study on Utica-Point Pleasant Shale Play Alireza Shahkarami and Guochang Wang 7.1 Introduction 7.2 Methodology 7.2.1 The Base Reservoir Simulation Model 7.3 Optimization Scenarios 7.4 Results and Discussion 7.4.1 Base Reservoir Model – A Single Well Case 7.4.2 Multi-Lateral Depletion – Finding the Optimum Number of Wells 7.4.3 Completion Parameters 7.4.4 Second Economic Scenario, Reducing the Cost of Completion 7.5 Conclusion Acknowledgments

110 114 120 121 122 122 125 126 127 128 129 130 131 131 134 136 139 139 139 141 142 143 143 147 148 148 148 151 153 154 156

Contents

Part 4: Fracture Reservoir Characterization Ahmed Ouenes Introduction References 8 Geomechanical Modeling of Fault Systems Using the Material Point Method – Application to the Estimation of Induced Seismicity Potential to Bolster Hydraulic Fracturing Social License Nicholas M. Umholtz and Ahmed Ouenes 8.1 Introduction 8.2 The Social License to Operate (SLO) 8.3 Regional Faults in Oklahoma, USA and Alberta, Canada used as Input in Geomechanical Modeling 8.4 Modeling Earthquake Potential using Numerical Material Models 8.5 A New Workflow for Estimating Induced Seismicity Potential and its Application to Oklahoma and Alberta 8.6 The Benefits of a Large Scale Predictive Model and Future Research 8.7 Conflict of Interest Acknowledgements References 9 Correlating Pressure with Microseismic to Understand Fluid-Reservoir Interactions During Hydraulic Fracturing Debotyam Maity 9.1 Introduction 9.2 Method 9.2.1 Pressure Data Analysis 9.2.2 Microseismic Data Analysis 9.3 Data 9.4 Results 9.4.1 Pitfalls in Analysis 9.5 Conclusions 9.6 Acknowledgements References

ix

159 159 161

163 164 165 166 168 173 178 179 179 179 181 181 182 182 186 187 188 196 196 197 197

x

Contents

10 Multigrid Fracture Stimulated Reservoir Volume Mapping Coupled with a Novel Mathematical Optimization Approach to Shale Reservoir Well and Fracture Design Ahmed Alzahabi, Noah Berlow, M.Y. Soliman and Ghazi AlQahtani 10.1 Introduction 10.2 Problem Definition and Modeling 10.2.1 Geometric Interpretation 10.2.1.1 Fracture Geometry 10.2.2 The Developed Model Flow Chart 10.2.3 Well and Fracture Design Vector Components 10.3 Development of a New Mathematical Model 10.3.1 Methodology 10.3.2 Objective Function 10.3.3 Assumptions and Constraints Considered in the Mathematical Model 10.3.3.1 Sets 10.3.3.2 Variables 10.3.3.3 Decision Variables 10.3.3.4 Extended Sets 10.3.3.5 Constant Parameters 10.3.3.6 Constraints 10.3.4 Stimulated Reservoir Volume Representation 10.3.5 Optimization Procedure 10.4 Model Building 10.4.1 Simulation Model of Well Pad and SRV’s Evaluation 10.5 Results and Discussions 10.6 Conclusions and Recommendations References Appendix A: Abbreviations Appendix B: Definition of the Fracturability Index Used in the Well Placement Process Appendix C: Geometric Interpretation of Parameters Used in Building the Model

199

200 203 203 203 204 204 204 207 207 207 208 208 208 208 209 209 210 211 212 214 216 216 218 220 220 221

11 A Semi-Analytical Model for Predicting Productivity of Refractured Oil Wells with Uniformly Distributed Radial Fractures 227 Xiao Cai, Boyun Guo and Gao Li 11.1 Introduction 228 11.2 Mathematical Model 229 11.3 Model Verification 231

Contents 11.4 Sensitivity Analysis 11.5 Conclusions Acknowledgements References Appendix A: Derivation of Inflow Equation for Wells with Radial Fractures under Pseudo-Steady State Flow Conditions

xi 231 233 234 234

235

Part 5: Environmental Issues of Hydraulic Fracturing 243 Introduction References 12 The Role of Human Factors Considerations and Safety Culture in the Safety of Hydraulic Fracturing (Fracking) Jamie Heinecke, Nima Jabbari and Najmedin Meshkati 12.1 Introduction 12.2 Benefits of Hydraulic Fracturing 12.3 Common Criticisms 12.4 Different Steps of Hydraulic Fracturing and Proposed Human Factors Considerations 12.5 Hydraulic Fracturing Process: Drilling 12.6 Hydraulic Fracturing Process: Fluid Injection 12.7 Fracking Fluid 12.8 Wastewater 12.9 Human Factors and Safety Culture Considerations 12.9.1 Human Factors 12.9.1.1 Microergonomics 12.9.1.2 Macroergonomics 12.9.2 Safety Culture 12.10 Examples of Recent Incidents 12.11 Conclusion and Recommendations Acknowledgment References 13 Flowback of Fracturing Fluids with Upgraded Visualization of Hydraulic Fractures and Its Implications on Overall Well Performance Khush Desai and Fred Aminzadeh 13.1 Introduction 13.2 Assumptions 13.3 Upgraded Visualization of Hydraulic Fracturing

243 245 247 248 250 250 252 254 257 258 258 259 259 260 260 261 263 265 266 266

271 272 272 273

xii

Contents

13.4

13.5 13.6 13.7 13.8

13.3.1 Concept 13.3.2 Results Reasons for Partial Flowback 13.4.1 Fracture Modelling 13.4.2 Depth of Penetration 13.4.3 Closing of Fractures 13.4.4 Chemical Interaction of Fracturing Fluids Impact of Parameters under Control Loss in Incremental Oil Production Conclusions Limitations References Appendix A

14 Assessing the Groundwater Contamination Potential from a Well in a Hydraulic Fracturing Operation Nima Jabbari, Fred Aminzadeh and Felipe P. J. de Barros 14.1 Introduction 14.2 Risk Pathways to the Shallow Groundwater 14.3 Problem Statement 14.4 Mathematical Formulation 14.5 Hypothetical Case Description and the Numerical Method 14.6 Results and Discussion 14.7 Conclusion References

Index

273 274 275 275 276 277 277 278 279 280 281 281 282 285 286 288 289 290 291 294 297 298

303

Foreword Since its inception in 1947, hydraulic fracturing has had a greater influence on hydrocarbon production than any other technology. Drilling and fracturing of horizontal wells starting in the late 1980’s enabled further increases in production. The oil industry took the new technology to a new level, creating the shale revolution, making the USA the world’s largest producer of crude oil. The advancement of technology also brought new challenges in technological, environmental, human, and geopolitical dimensions. Each challenge will have to be understood and addressed. This book collects papers that address the environmental and human dimensions in simple terms in order to clarify issues and dispel misconceptions. Exploring the technological dimension, the book includes chapters that address many of the issues that are very unique to unconventional reservoirs, including chapters on Reservoir-Fluid interaction, optimization of fracture spacing, and optimization of fracture placement.

M. Y. Soliman, PhD, PE, NAI Chair of the Petroleum Engineering Department William C Miller Endowed Chair The University of Houston

xiii

Part 1 INTRODUCTION

Fred Aminzadeh (ed.) Hydraulic Fracturing and Well Stimulation, (1–33) © 2019 Scrivener Publishing LLC

1 Hydraulic Fracturing, An Overview* Fred Aminzadeh

*

University of Southern California

Abstract This article provides an overview of the state of the art in hydraulic fracturing, a

controversial topic of the last decade. To Frack or not to Frack, That is the Question; was posed at a meeting of the Western Regional Society of Petroleum Engineers. The fact is, we have witnessed an intense debate over hydraulic fracturing’s economic benefits and the role it has played in securing energy independence, and its ill effects (perceived or real) during the past decade. Many, in particular those in the fossil energy industry, consider shale oil/gas, with the associated horizontal drilling and hydraulic fracturing, as one of the major developments in the oil and gas industry of the past two decades. Others, especially many environmentalists, consider fracking proponents as public enemy number one. Different sections of this entry attempt to highlight different scientific facts about hydraulic fracturing, the common-sense environmental concerns, and the respective economic ramifications. After a brief overview of the principles of hydraulic fracturing in section “What Is Hydraulic Fracturing?”, we discuss the importance of hydraulic fracturing in section “Why Hydraulic Fracturing Is Important.” This is followed by fracture characterization (section “Fracture Characterization”) and geomechanics (section “Geomechanics of Hydraulic Fracturing”). They examine the natural fractures in the subsurface and how one can characterize them, how hydraulic fracturing helps to expand the natural fractures and/or create new (stimulated) fractures as well as the underlining rock mechanics properties and the related stress regime. Section “Environmental Aspects of Hydraulic Fracturing” addresses different environmental concerns about hydraulic fracturing. This includes the potential ground water contamination, amount of water used for hydraulic fracturing the water disposal process, and methane emission concerns. Another environmental concern is the risk of induced seismicity or man-made earthquakes. Given the Email: [email protected] *Adopted from Aminzadeh, F., 2018, Hydraulic Fracturing, An Overview, Journal of Sustainable Energy Engineering, Vol. 6, Issue 4, pp. 204–228. Fred Aminzadeh (ed.) Hydraulic Fracturing and Well Stimulation, (3–33) © 2019 Scrivener Publishing LLC

3

4

Hydraulic Fracturing and Well Stimulation

publicized controversy on whether and to what extent hydraulic fracturing creates induced seismicity section “Induced Seismicity” is dedicated to deal with this issue. The key message of this article is the best way to answer the question with which we began, namely to frack or not to frack, lies with science; the hope is, with sound scientific and engineering investigation, truth will prevail - “veritas omnia vincit.” Keywords: Hydraulic fracturing (HF), shale resources, energy resources, micro-seismic data, economic factors, environmental impact of HF, fracture characterization, geomechanics, induced seismicity, ground water contamination, stress field

1.1 What is Hydraulic Fracturing? Hydraulic fracturing (HF) is an oil and gas operation used to recover hydrocarbon resources that are trapped in low-permeability shale and other lithologies. Over geologic time periods, these resources were formed by the maturation of kerogen, the organic precursor of petroleum. However, unlike conventional oil and gas where hydrocarbons migrate into the reservoir from a separate source, the hydrocarbon source and reservoir rock share low permeability, forming an unconventional system. While there are significant volumes of hydrocarbon trapped in unconventional reservoirs, the extremely low natural permeability of the formations impedes commercial production using conventional techniques. HF is a process that involves injection of large volumes of water (several million gallons), sand, and small volumes of chemical additives to increase oil or natural gas flow from low permeability formations. The large pressure associated with injection of “fracturing fluid” creates new fractures and extends existing fractures that enhance hydrocarbon flow, while sand mixed with injected fluid holds the new and existing fractures open. Some of the injected fluid flows back to the wellbore and is pumped to the surface, or is injected back to the reservoir. Figure 1.1 is an example of a typical HF configuration. HF is usually performed on horizontal or directional wells, oftentimes with the well azimuth being perpendicular to the direction of maximum horizontal in situ stress. Figure 1.1 shows a schematic display of the horizontal well and the enlarged shale fractures are shown in the bottom left part of Figure 1.1 display of the horizontal well and the enlarged shale fractures are shown in the top right with the multistage (3 stages) fracturing. Although HF has been used since the 1950s, over the last decade it has been the subject of intense public debate. Some of the concern has been over its potential impacts on drinking water, the potential for emission of

Hydraulic Fracturing, an Overview USDW

5

Private well Municipal water well: 0, that corresponds to a tensile change in the effective normal traction. In the current model, we neglect hydraulic fracture propagation and fluid leak-off, and we focus on early-time interaction between hydraulic and natural fractures. In particular, we quantify the evolution of state

Evolution of Stress Transfer Mechanisms 43 of stress and slip on the natural fractures as a function of opening of the hydraulic fracture. We use the standard finite element method with bi-linear elements in the bulk and linear interface elements on the fractures to discretize Eqns. (2.6–2.7).

2.4 Numerical Model We solve the quasi-static mechanical equilibrium problem and the fracture contact problem, Eqns. (2.6–2.7), using PyLith [20] to calculate displacements and stresses in the 2D domain and slip and tractions on the 1D fractures. We use the finite element grid shown in Figure 2.1 with Lx = 200 m and Ly = 100 m and a 100 m long hydraulic fracture. We use triangular elements in the matrix and linear interface elements along the fractures. We consider a simulation with the following values: σ1 = σxx (Lx,y,0) = –40 MPa and σ3 = σyy (x,Ly,0) = –30 MPa as the boundary tractions, and p0 = p(x,y,0) = 10 MPa as the initial pressure. We assume E = 56.82 GPa and ν = 0.26. We use a static friction model with τc = 0 and μf = 0.1, which, for the values chosen above for σ1 and σ3, results in shear slip on the natural fractures as soon as hydraulic fracturing begins. We assume that water is injected in the center of HF at a constant bottomhole overpressure of δpf(Lx/2,Ly/2,t) = 44 MPa. The other parameters are as follows: water compressibility of 1.45E-08 per Pa, water viscosity of 1 cP, HF permeability of 1 darcy, and HF porosity of 0.5. We calculate the evolution of HF pressure δpf(x,Ly/2,t) along the length of HF at discrete time steps (Figure 2.2).

50 t

120 sec

pf, Mpa

40 30

1 sec

20 10 0 50

100 x, m

150

Figure 2.2 Time evolution of overpressure along HF calculated using the analytical solution of the fluid mass balance equation for a constant pressure injection in a porous medium.

44

Hydraulic Fracturing and Well Stimulation

A non-uniform time stepping is chosen with finer resolution in the beginning to resolve the early time transient and coarser resolution later as the pressure reaches a pseudo steady state value. We conduct the simulation for 0 < t < 121 sec. Initial stresses balance boundary tractions ensuring zero initial displacements.

2.5 Simulation Results Opening of the hydraulic fracture with time leads to an evolution of the state of stress and resulting deformation in the domain, which drives the time-dependent response of the natural fractures. Overpressure causes opening of HF resulting in uy > 0 above HF and uy < 0 below HF (Figures 2.3 and 2.4). Opening is asymmetric in the y-direction because the top boundary is a compression boundary and the bottom boundary is a fixed ux 100 0

y, m

meter –0.07

0 0 (a)

–0.035 –0.035 x, m

200 uy

0.0

0.035

0.07 (b)

Figure 2.3 Displacement field at t = 121 sec resulting from a balance of boundary conditions and hydraulic fracturing induced stress changes. The horizontal displacement, ux, changes sign as we move from HF to the top and bottom boundaries of the domain. Its magnitude is asymmetric along x due to asymmetry in the right and left boundary conditions. Magnitude of the vertical displacement, uy, is asymmetric along y because of the asymmetry in the top and bottom boundary conditions. For the ease of visualization, the mesh is distorted with displacements magnified by a factor of 50, and the color scale is chosen to show the variation in ux and uy and not the minimum and maximum values of these quantities, which are –0.022 m and 0.026 m for and –0.018 m and 0.092 m for uy. The dash lines in the upper figure indicate the location of the profile plots in Figure 2.4.

y, m

Evolution of Stress Transfer Mechanisms 45 100

100

80

80

60

60

40

40 xx

ux uy

20 0

–0.05

0

0.05 0 Displacement, m

–20

0 MPa

20

20

0.04

0

–0.04

0 (c)

yy xy

((b)

MPa

Displacement, m

(a))

20

0

–20 50

100 x, m

150

200

0 (d)

50

100 x, m

150

200

Figure 2.4 Profiles of displacements and stress changes orthogonal (upper row) and parallel (lower row) to HF at t = 121 sec. Orthogonal profiles are at x = 70 m and parallel profiles are at y = 25 m along dash lines shown in Figure 2.3a. (a) Displacement discontinuity across HF is visible. ux changes sign above HF (y > 50 m) and similar to xx . uy is negative below HF and positive above HF, as expected. (b) The vertical stress change, yy , is compressive due to opening of HF. The horizontal stress change, xx , is also compressive in the immediate vicinity of HF but becomes tensile as we approach the traction boundary on the top because of upward displacement. The shear stress change, xy, due to HF opening is larger above HF because of a larger upward displacement of the upper surface of HF. (c) Across the center of HF, ux is asymmetric due to the asymmetry in the horizontal boundary conditions, and uy is symmetric. Both displacements have discontinuities at x = 25 m and x = 175 m where we cross NF2 and NF4, respectively. Stresses are piecewise continuous because we use linear displacement elements.

displacement boundary. Displacement in the x-direction, ux, reflects the stress field generated by the balance of boundary tractions, fixed displacement boundaries, and pore pressure. The compressional boundaries on top and right, with the principal stress ratio of σ1/σ3 = 1.5, and the fixed boundaries on bottom and left cause ux > 0 on the left half of the fracture surface and ux < 0 on the right half of the fracture surface.

46

2.6

Hydraulic Fracturing and Well Stimulation

Effect of Hydraulic Fracturing on Natural Fractures

The initial values of tractions on NF can be computed analytically using Cauchy’s formula. We know that the effective normal and shear traction on the fracture surfaces are given as follows [8]

N

1 2

xx

1 2

yy

1 2

xx

xx

yy

cos 2

pf (2.8)

yy

sin2

This results in the following initial values: N (t 0) 20 MPa MPa and τ(t = 0) = 0 on HF, N (t 0) 25 MPa and τ(t = 0) = 5 MPa on NF1 and NF4, which have their normal vectors such that α = 135 degree, and 25 MPa and τ(t = 0) = –5 MPa on NF2 and NF3, which have N (t 0) their normal vectors such that α = 45 degree. The ratio of shear to effective normal traction of 0.2 at t = 0 is higher than the static friction of μf = 0.1 resulting in slip on natural fractures. Traction and slip on the natural fractures evolve dynamically as the hydraulic fracture opens under a time-dependent pressure increase from injection. To understand the evolution of tractions on the natural fracture surfaces, we analyze the evolution of stresses in elements adjoining the midpoints of the natural fractures as shown in Figure 2.5. Opening of the hydraulic fracture causes sliding of the natural fractures (Figure 2.6a) even in absence of hydraulic communication. This is reported as dry microseismic events in the literature [7]. The sliding direction depends on the relative angle α between a natural fracture and the maximum principal stress orientation. The shear traction magnitudes on NF1, NF2, and NF3 decrease due to slip-induced relaxation (Figure 2.6c). The effective normal compressions on NF1, NF2, and NF3 decrease because opening of HF leads to vertical expansion of the domain and tensile changes around the natural fractures (Figure 2.6d). NF4 behaves differently from the other three natural fractures due to its position in the domain close to the maximum principal stress boundary and the fixed bottom boundary. The shear traction on NF4 drops from 5 MPa to 2.48 MPa at the first time step and remains constant as it slips. As explained in Figure 2.5 caption, the effective compression on NF4 increases causing a frictional stabilization away from the failure line in the stress space (Figure 2.7a).

Evolution of Stress Transfer Mechanisms 47

σxy

MPa

0

–10 σ΄yy

NF1 NF2 NF3 NF4

–20 σ΄xx –30

0

100

50 Time, sec

Figure 2.5 Time evolution of the effective stresses with opening of the hydraulic fracture. The evolution is plotted for elements near the midpoints of the four natural fractures. All natural fractures experience a decrease in compression and an increase in shear magnitude due to opening of HF. NF4 experiences a relatively smaller decrease in compression and largest increase in shear, which results in a compressive change in its normal traction compared to tensile changes observed on the other three natural fractures (Figure 2.6). 0.01 NF1 NF2 NF3 NF4

0

–0.01 0 (a)

NF dnormal, m

NF dshear, m

0.01

0

–0.01 0 (b)

100 50 Time , sec

50

100

–20

5 NF σ΄n, MPa

NF τ, MPa

–22 0

–5 0 (c)

50

100

–24 –26 –28 –300 (d)

50

100

Figure 2.6 Time evolution of slip (top row) and traction (bottom row) at the midpoints of the four natural fractures during hydraulic fracturing. (a) Positive shear values on NF1 and NF4 indicate left-lateral shear and negative shear values on NF2 and NF3 indicate right-lateral shear. (b) Normal slip values are zero indicating no opening or closing of the natural fractures. (c) Shear traction magnitude decreases for all natural fractures as they slip. This decrease continues longer for NF1, NF2, and NF3 than for NF4. (d) Normal compression decreases on NF1, NF2, and NF3 thereby destabilizing them. It increases on NF4, which has a stabilizing effect. Positive values indicate opening or tension and negative values indicate closing or compression of the fractures.

48

Hydraulic Fracturing and Well Stimulation 3

0.01 Time

Time

1

NF1 NF2 NF3 NF4

0

–1

0

e Tim

–2 –3 –30 (a)

–0.01 –25

–20

0

NF σ΄n, MPa

(b)

3

–20

2

–22 NF σ΄n, MPa

NFτ, MPa

NF dshear, m

NFτ, MPa

2

1 0

–1

(c)

0.1

–24 –26 –28

–2 –3 0

0.05

HF center dnormal, m

20 HF Ptip, MPa

40

–30 0 (d)

20 HF Ptip, MPa

40

Figure 2.7 Evolution of traction and slip on natural fractures due to opening and pressurization of the hydraulic fracture. (a) Stress paths of tractions at the midpoints of the four natural fractures showing that NF1 and NF4 start sliding under positive (leftlateral) shear tractions whereas NF2 and NF3 start sliding under negative (right-lateral) shear tractions. The dash lines are the failure lines at μf = 0.1. The initial stress state ( N , )t 0 ( 25, 5) MPa falls outside the chosen scales of the plot. (b) Shear slips of natural fractures increase monotonically with the opening at the center of HF. (c) Shear tractions on the natural fractures decrease monotonically with the HF tip pressure. (d) Effective normal compressions decrease on NF1, NF2, and NF3 and increase on NF4 with increasing HF tip pressure.

The stimulated reservoir volume (SRV) can be quantified in terms of the energy that is transferred from the hydraulic fracture to natural fractures (Figure 2.8a). We observe in our simulation that opening of HF transfers elastic energy to NF1, NF2, NF3, and NF4, where the energy per unit fracture surface area is calculated as l·d. The fastest rate of energy transfer is observed for NF4 that experiences highest amount of slip and compression during hydraulic fracturing. Microseismicity of natural fractures can also be related quantitatively to hydraulic fracturing (Figure 2.8b). We calculate 2 the moment magnitude as, M w log10 M0 6.07 where M0 G | d | d 3 f

0.03

0.02

0.01

0

0

(a)

2

NF1 NF2 NF3 NF4 4

HF strain energy/area, MPa

Moment magnitude , Mω

NF strain energy/area, MPa

Evolution of Stress Transfer Mechanisms 49 0.5 0.4 0.3 0.2 0.1 0

0

(b)

0.05

0.1

HF center dnormal, m

Figure 2.8 (a) Energy transfer from the hydraulic fracture to the natural fractures is monotonic. Growth in NF2 and NF4 strain energy are larger due to larger shear traction and slip near the fixed bottom boundary. (b) The seismic moment magnitudes of natural fracture slip events evolve with the opening of hydraulic fracture. Natural fractures are grouped based on their position with respect to the traction boundary in the minimum principal stress direction.

is the seismic moment, G is the shear modulus, and |d| is the slip magnitude [22]. We find that natural fractures are grouped by their proximity to the traction boundary in the minimum principal stress direction i.e. NF2 and NF4, which are farther from the top boundary, are in a group with a faster increase in the moment magnitude compared to NF1 and NF3, which are closer to the top boundary.

2.7 Conclusion We investigated the effect of hydraulic fracturing on pre-existing natural fractures by analyzing the time-evolution of tractions and slips of natural fractures with opening of the hydraulic fracture. We observe that the profiles of normal and shear slip along the natural fractures depend on the relative position of the natural fracture with respect to the hydraulic fracture, the fixed displacement boundaries, and the minimum and maximum principal stress orientations. We explain the heterogeneity in slip response of the natural fractures in terms of the displacement and stress profiles parallel and orthogonal to the hydraulic fracture. The energy transfer approach in our hydraulic fracture–natural fracture interaction model can be used to capture slip-induced permeability enhancement and microseismic response of natural fractures during field scale simulations of hydraulic fracturing. We are enhancing our framework to allow for two-way coupling

50

Hydraulic Fracturing and Well Stimulation

among the processes of fluid flow, deformation, and seismicity. We are also implementing the capability to allow fracture propagation. The proposed model allows for improved prediction of stimulated reservoir volume and induced seismicity during hydraulic stimulation.

References 1. Hubbert, M.K. and Willis, D.G., Mechanics of hydraulic fracturing. Trans. Am. I. Min. Met. Eng., 210, 6, 153–163, 1957. 2. Hossain, M.M., Rahman, M.K., Rahman, S.S., A shear dilation stimulation models for production enhancement from naturally fractured reservoirs. SPE J., 7, 2, 183–195, 2002, doi: 10.2118/78355-PA. SPE-78355-PA. 3. Curtis, J.B., Fractured shale-gas systems. Am. Assoc. Petrol. Geologists Bull., 86, 1921–1938, 2002. 4. Lamont, G.C. and Jassen, F., The effects of existing fractures in rocks on the extension of hydraulic fractures. J. Petrol. Technol., 15, 203–209, 1963. 5. Blanton, T.L., An experimental study of interaction between hydraulically induced and pre-existing fractures. Paper SPE 10847 presented at the SPE/ DOE unconventional gas recovery symposium, Pittsburg, Pennsylvania, May, 1982. 6. Huang, J., Safari, R., Mutlu, U., Burns, K., Geldmacher, I., McClure, M., Jackson, S., Natural-hydraulic fracture interaction: Microseismic observations and geomechanical predictions. unconventional resources technology conference (URTeC), Denver, Colorado, USA, 25–27 August, 2014, doi: 10.15530/urtec-2014-1921503. 7. Nagel, N.B., Sanchez-Nagel, M.A., Zhang, F., Garcia, X., Lee, B., Coupled numerical evaluations of the geomechanical interactions between a hydraulic fracture stimulation and a natural fracture system in Shale Formations. Rock Mech. Rock Eng., 46, 581–609, 2013, doi: 10.1007/s00603-013-0391-x. 8. Jaeger, J.C., Cook, N.G.W., Zimmerman, R.W., Fundamentals of Rock Mechanics, fourth edition, Blackwell Publishing, Malden, MA, 2007. 9. Olson, J.E., Bahorich, B., Holder, J., Examining hydraulic fracture-natural fracture interaction in hydrostone block experiments. SPE 15261, Society of Petroleum Engineers, 2012, doi: 10.2118/152618-MS. 10. Kresse, O., Weng, X., Gu, H., Wu, R., Numerical modeling of hydraulic fractures interaction in complex naturally fractured formations. Rock Mech. Rock Eng., 46, 555–568, 2013, doi: 10.1007/s00603-012-0359-2. 11. Warpinski, N.R. and Teufel, L.W., Influence of geologic discontinuities on hydraulic fracture propagation. J. Petrol. Tech., 209–220, 1987. 12. Zangeneh, N., Eberhardt, E., Bustin, R.M., Investigation of the influence of natural fractures and in situ stress on hydraulic fracture propagation using a distinct—element approach. Can. Geotech. J., 52, 926–946, 2015.

Evolution of Stress Transfer Mechanisms 51 13. Aimene, Y.E. and Ouenes, A., Geomechanical modeling of hydraulic fractures interacting with natural fractures-Validation with microseismic and tracer data from the Marcellus and Eagle Ford. Interpretation-J. Sub., 3, 71–88, 2015, doi: 10.1190/INT-2014-0274.1. 14. Potluri, N., Zhu, D., Hill, A.D., Effect of natural fractures on hydraulic fracture propagation. SPE European Formation Damage Conference, The Netherlands, May 2005, 2005, SPE 94568. 15. Gu, H., Weng, X., Lund, J., Mack, M., Ganguly, U., Suarez-Rivera, R., Hydraulic fracture crossing natural fracture at nonorthogonal angles: A criterion and its validation. SPE Hydrualic Fracturing Technology Conference and Exhibition, The Woodlands, Texas, 24–26 January 2011, 2011, SPE 139984. 16. Nassir, M., Settari, A., Wan, R., Prediction of stimulated reservoir volume and optimization of fracturing in tight gas and Shale with a fully elasto-plastic coupled geomechanical model. SPE J., 19, 5, 771–785, 2014. 17. Cheng, W., Jin, Y., Chen, M., Reactivation mechanism of natural fractures by hydraulic fracturing in naturally fractured shale reservoirs. J. Nat. Gas Sci. Eng., 23, 431–439, 2015. 18. Ouenes, A., Aissa, B., Boukhelf, D., Fackler, M., Estimation of stimulated reservoir volume using the concept of Shale Capacity and its validation with microseismic and well performance: Application to the Marcellus and Haynesville. SPE Western North American and Rocky Mountain Joint Regional Meeting held in Denver, Colorado, USA, 16–18 April 2014, 2014, SPE 169564. 19. Jha, B. and Juanes, R., Coupled multiphase flow and poromechanics: A computational model of pore pressure effects on fault slip and earthquake triggering. Water Resour. Res., 50, 3776–3808, 2014, doi: 10.1002/2013WR015175. 20. Aagaard, B.T., Knepley, M.G., Williams, C.A., A domain decomposition approach to implementing fault slip in finite-element models of quasi-static and dynamic crustal deformation. J. Geophys. Res. Solid Earth, 118, 3059– 3079, 2013, doi: 10.1002/jgrb.50217. 21. Carslaw, H.S. and Jaeger, J.C., Conduction of Heat in Solids, second edition, Clarendon Press, Oxford, 1959. 22. Hanks, T.C. and Kanamori, H., A moment magnitude scale. J. Geophys. Res., 84, 2348–2350, 1979.

3 Primer on Hydraulic Fracturing Concerning Initiatives on Energy Sustainability Michael Holloway1* and Oliver Rudd2 1

NCH Corporation, Irving, TX, USA 2 Independent Consultant, TX, USA

Abstract Hydraulic fracturing (also known as fracking, fracing, and other variations) of rock deep beneath the Earth’s surface to release petroleum product has become a contentious subject across the globe. This practice is not to be confused with drilling or extraction. Fracing is the process of using fluid power to fracture rock to release gas and sometimes crude oil. It is not drilling, although drilling must be done to establish a well in order to pump fluid thus fracturing rock to release product. Certain countries have actually outlawed the practice of hydraulic fracturing claiming that ground water and air pollution increase due to the practice. There is also the claim that the comfort of life is adversely affected. Legitimate concerns are always available and examples to purport a concerned view are magnified, in the authors’ opinion. The intent of this paper is to provide a correct and balanced view of fracturing underground rock with fluids in order to release a product to produce energy. The concept of using water to do work is nothing new. Pumping fluid below ground in order to fracture rock to release gaseous petroleum is, however, a relatively new practice. It is done with surprising precision as well as environmental concern yet it is interesting how the public reacts to the practice in relation to other techniques used throughout the world. This paper will explore the materials used as well as the concerns most common to the practice.

*Corresponding author: [email protected] Fred Aminzadeh (ed.) Hydraulic Fracturing and Well Stimulation, (53–78) © 2019 Scrivener Publishing LLC

53

54

Hydraulic Fracturing and Well Stimulation

3.1 Hydraulic Fracturing When people say that you cannot squeeze blood out of a turnip, it means that you cannot get something from a person, especially money that they do not have. You cannot squeeze blood from a turnip but you can release trapped natural gas from rock – at least that is what is being accomplished now throughout North America. One hundred years ago no one thought it possible. Very few if any contemplated the idea. Natural gas, which is primarily methane, has been proven to be an excellent fuel source. It can be safely burned to create heat to power engines, boilers in factories and homes as well as powering turbines for generating electricity. Projections on natural gas volumes trapped under ground suggest a near inexhaustible supply of this product yet with such abundance spawns controversy. A popular and economical technique relies on the gas from subterranean sources requires fracturing rock bed. This process is actually carried out naturally everyday with water or magma. Magma may flow into rock beds, superheating water to generate steam. The resulting pressure of the expanding water molecule can be so great that it can lift and separate thousands of tons of rock deep beneath the Earth’s surface. This same practice can be carried out artificially (induced) using high powered pumps and various liquid compounds. This technique, combined with new horizontal directional drilling machines, has enabled the harvest and distribution of natural gas. But at what cost? Does this practice contribute to greenhouse gas? Does it create earthquakes? Does it contaminate the ground water supply? These are important ideas to consider yet, and with proper examination and logic, we are confident you will gain insight and reason in a practice fueled by profit and civil concern. No sides are taken in the following pages. This work is not intended to be pro-industry or anti-fracturing. Namely, it aims to educate the general public on what hydraulic fracturing really is, how it is conducted and what possible harms may or may not come as a result.

3.1.1 Environmental Impact – Reality vs. Myth In today’s society, it is really easy for organizations – be it the general media, political groups, local organizations, unions or religious associations – to spread their beliefs to the public and push whatever agenda or ideals they may have. These beliefs could be successfully put forward with good intentions, successfully put forth with bad intentions or, in many cases, put forward with good intentions but have a negative result. Sadly,

Primer on Hydraulic Fracturing 55 it seems human nature dictates that the first opinion heard or the opinion heard the loudest and with the most hyperbole will be what the public comes to believe. In time, once something is believed by enough people and stated as “fact” long enough, the general public will no longer even bother looking into facts and it will become part of the fabric of beliefs in our society – for instance, a few examples of this phenomenon are: 1) you, in fact, cannot see The Great Wall of China from the moon (not even close); 2) the Sherlock Holmes character never once said “Elementary, my dear Watson”; and 3)  Neil Armstrong actually said, “One small step for [a] man, one giant leap for mankind.” As far as hydraulic fracturing is concerned, the aspect given the most attention by press and most all concerned organizations is the impact it may have on the environment. The question of environmental impact through fracking is, to say the least, a very emotional topic and by far the most polarizing issue; however, a great deal of analysis indicates that the most significant environmental risks attributed to fracking are similar to risks long associated with all drilling operations – including groundwater contamination due to inadequate cementing and/or well construction, risks associated with trucking, leaks from tanks and piping and spills from waste handling. This all-encompassing blame has given industry all the ammunition needed to claim effects attributed to hydraulic fracturing are overstated, not based on good science, or related to processes other than hydraulic fracturing. Due to the great ongoing controversy over alleged impacts from fracking, many public groups have become deeply suspicious of the trustworthiness and overall motives of the oil and gas industry. These suspicions are continuously intensified by two things: 1. ongoing mistrust of data and findings due, in great part, to semantics, and 2. by the industry initially refusing to disclose the chemical makeup of fracking fluids and the additives used to enhance hydraulic fracturing.

3.1.2 The Tower of Babel and How it Could be the Cause of Much of the Fracking Debate Most everyone has heard the story, or has a general understanding, of the Tower of Babel from the Old Testament. In the Biblical account of this story, humanity was attempting, as a unified group, to build a tower in

56

Hydraulic Fracturing and Well Stimulation

Mesopotamia to reach the heavens…only to have their efforts brought to a halt by one of the most effective means imaginable. The efforts of this united group of people were not thwarted by military force, or by weather, or even by sickness and injury…their efforts in this undertaking were thwarted by speech. The simple fact of this story is that building of the tower came to a halt once the unified people were confounded by speech and no longer able to communicate to work together. Now far be it for this work to compare modern day hydraulic fracturing with the construction, and subsequent stop in construction, of the Tower of Babel, but much of the confusion, name-calling and general mistrust between groups on this subject can be attributed to a difference in communication. Maybe once this communication gap is bridged more effective talks can be established between industry and concerned public – in place of wasting time on mistrust and name-calling. Hopefully, this work can help to bridge that gap. It can be easily considered that a very large portion of negativity toward hydraulic fracturing is actually attributable to processes other than hydraulic fracturing. In the discussions between industry and the public, a great deal of this problem can boil down to an issue of semantics…the oil and gas industry has a narrow view of what fracking entails (including just those processes related to the actual process of fracking while on location conducting the fracking operation) while the general public is more inclined to include many more activities commonly related to fracking (water and sand trucking, product and equipment transport and storage, water disposal), under the heading of “fracking.” This can cause misunderstandings and skewed data in that many of the processes included by the general public are utilized in many, if not all, drilling practices and are hard to put solely under the heading of “fracking” when in actuality they could just as easily be under the heading “completions” or “production.” As has been discussed many times in the media and will undoubtedly be discussed again and again and again ad nauseam, there are many proven environmental impacts caused by drilling operations and processes related to drilling. This fact can almost assuredly not be disproven by industry personnel and can be a concern by the public in their feelings on gas well completion and production activities. However, by the same token, there have been well over a million wells that have gone through the process of hydraulic fracturing, as is defined by industry, with not one reported instance of ever impacting a fresh water aquifer. With the current issue of semantics, public concerns can include many drilling processes while industry can fall back on the fact that the industry definition of fracking has never impacted fresh water in the ways commonly claimed by the media for public consumption…and the debate can rage on with both sides being

Primer on Hydraulic Fracturing 57 right and both sides being wrong while never taking steps to come together on a common goal.

3.1.3 Frac Fluids and Composition The use of hydraulic fracturing for oil and gas exploration in the US has become highly controversial with one of the greatest points of contention between the public and industry being the makeup of frac fluids and their possible impacts on public health and the environment. This has become such a hot topic with many segments of the public for two reasons: 1. if a concern exists about the pumping of fluids into any structure, then the most concern will naturally be centered on what is being pumped, and 2. a great deal of suspicion arose and were intensified when the oil and gas industry initially balked at the disclosing the chemical makeup of fluids used to enhance hydraulic fracturing. This has become a major argument point for the concerned public because, basically, “if you have nothing to hide why would you not want to disclose it?” Take, for example, a small child walks into a room with his hands behind his back and will not show you what is there for several minutes…and then only does when forced. Well, even if it turns out he was holding something as harmless as a feather behind his back it will make you suspicious all the same wondering “what was he doing with that feather!” To make this contentious subject a little clearer, this section will provide descriptions of why frac fluids are needed, what general chemicals are needed and used, relative amounts of chemicals in frac fluid composition, proppants – the different types and uses, a discussion on slickwater, and a discussion on present regulations and standards for industry disclosure of frac fluid compositions.

3.1.4 Uses and Needs for Frac Fluids There are a great deal of varied chemicals used every day in oil production wells during all phases of drilling, completions and production. These chemicals can include cements used to seal the annulus to protect the pipe and surrounding formation from damage through wells exceeding the producing and stimulation requirements placed on the pipe, temperature, and

58

Hydraulic Fracturing and Well Stimulation

even natural ground stresses. An example is corrosion inhibitors. These chemicals help pipe and connection seals remain within design specifications to prevent failures. Corrosion prevention and treating chemicals may also be necessary due to operational and field changes – even after well completion and during production. These chemicals can be used in much the same fashion as fracturing; however, chemicals in well operation are applied in smaller quantities, at lower pressure and in a regular maintenance driven schedule during a well’s life. Just like the maintenance driven chemicals utilized during operations, chemicals serve numerous necessary functions to insure successful, safe and efficient hydraulic fracturing operations. The following provides a comprehensive look at common chemical additives utilized in the current fracturing industry.

3.1.5 Common Fracturing Additives First, there is no one formula for how much each of the following additives are used in a given fracturing fluid; however, the following section is intended to present a brief description of some of the most commonly used additives and a general percentage breakdown of each that has been widely reported and is, therefore, easily verified by anyone wishing to do so. Also, each well differs in the number, type and amount of additives (please note: the term “additives” is used to include water, sand and chemicals to allow for a discussion of each under one heading) in a successful fracture treatment – “typically” between 3 and 12 additives depending on the conditions of the specific well to be fractured and characteristic of the surrounding formations. Additives utilized in hydraulic fracturing operations are intended to serve specifically engineered uses – such as biocides to control microorganism/bacterial growth, corrosion inhibitor to prevent corrosion of pipe, viscosity agents to carry proppant, gelling agents to improve proppant placement, friction reduction to decrease pump friction and reduce treating pressure, oxygen scavengers to also aid in corrosion prevention in metal pipes, and acids to help remove drilling mud buildup damage. Fluids (typically water) – usually approximately 98%–99% of the total volume – used to create the fractures in the formation and to carry a propping agent (typically silica sand) which is deposited in the induced hydraulic fractures to keep them from closing up. Hydrochloric acid (example is 15% HCl) – usually approximately 500 to 2,000 gallons per three thousand gallons of frac

Primer on Hydraulic Fracturing 59 fluid – used to help dissolve minerals and help remove damage near the well bore by cleaning out cement around pipe perforations, and also helps initiate fissures in the rock matrix. Corrosion inhibitor (example is ammonium bisulfate) – usually approximately 0.2%–0.5% of acid total volume, resulting in approximately 5–10 gallons – used only in instances when acid is used to prevent pipe corrosion. Biocides (examples are sodium hypochlorite or chlorine dioxide) – usually approximately 0.005%–0.05% of the total volume – used to control bacterial growth in the water injected into the well and prevent pipe corrosion. Friction reducers (examples are polyacrylamide based compounds) – usually approximately 0.025% of total volume – used to reduce pipe friction and pressure in the piping required to pump fluids. Gelling agents (examples are guar gum and cellulose) – not often used – used to thicken water-based solutions and help in suspension and transport of prop-pants into formation. Crosslinking agent (examples include boric acid, titanate and zirconium) – used to enhance abilities of the gelling agent to even further aid in transport of prop-pant material. Breaker solution – when cross-linking additives are added, a breaker solution is commonly added in the frac stage to cause the enhanced gelling agent to break down into a simpler fluid so it can be readily removed from the wellbore without carrying back the sand/proppant material. Oxygen scavenger (example includes ammonium bisulfate) – used to prevent corrosion of pipe by oxygen. Iron control and stabilizing agents (examples are citric acid and acetic acid) – used to keep iron compounds in soluble form to prevent precipitation. Surfactant – usually approximately 0.5 to 2 gallons per thousand gallons of frac fluid – is used to promote flow of the fluids used in the fracturing process. Scale Inhibitor (example is ethylene glycol) – seldom used – used to control the precipitation of specific carbonate and/or sulfate minerals. Proppants (examples are sand, resin coated sand or man-made ceramic particles) – usually approximately 1%–1.9% of total volume – used to hold fissures open so gas and oil can be extracted.

60

Hydraulic Fracturing and Well Stimulation

Now, I am sure many of you that have seen this type of information before are now expecting to see one of those “other uses” tables telling you that fracking fluid must be safe due to the “ingredients” of fracking fluids having such everyday uses such as: scale inhibitors having the same chemicals as windshield washer fluid, friction reducers having the same chemicals as many makeup products, surfactants being basically the same as shampoo products, proppants being play sand and hydrochloric acid also being swimming pool cleaner; these may be true in the strictest sense of the word, but this type of listing can also be very misleading and insincere, in that most all chemicals can be used for many different things, but are still not something with which you necessarily want to come in contact. For example, ammonium nitrate is commonly used in agriculture as a high-nitrogen fertilizer, nitromethane is a commonly used industrial solvent and Ryder trucks are commonly used to move families and their belongings to their dream homes – while these are also three of the common “ingredients” used in the tragic April 1995 Oklahoma City bombing of the Alfred P. Murrah Federal Building, which killed 168 people. This is, of course, a comparison made for shock value, but it is meant as such to stick in your memory as how these sorts of comparisons can be manipulated and to drive home the fact that the best policy is to study upon facts when you see a comparison like this, and make an informed decision for yourself.

3.1.6 Typical Percentages of Commonly Used Additives Fracturing fluids are varied to meet the specific needs of each location; however, evaluating the widely reported percentage volumes of the fracturing fluid components reveals the relatively small volume of additives that are present. Overall, the concentration of additives in most fracturing fluids is a relatively consistent 0.5% to 2%, with water and proppants making up the remaining 98% to 99.5%. Keep in mind, however, that a typical fracturing job uses upwards of five million gallons of fracturing fluid, so a small percentage amount may actually result in a great deal of chemical usage, no matter how diluted it may be. As you can imagine, the overall composition of fracturing fluids varies among companies and the drilling location. However, as a pretty good baseline, fracturing fluids typically contain: Approximately 90% water Approximately 9.5% proppant materials Approximately 0.5% chemicals – this percentage varies, but is typically between 0.5–1.0% by weight of total fluid

Primer on Hydraulic Fracturing 61

KCI 0.06% Surfactant 0.085% Water and sand 99.51%

Other 0.49%

Friction reducer 0.088%

Gelling agent Scale pH adjusting 0.056% inhibitor agent 0.043% 0.011% Breaker 0.01% Crosslinker 0.007% Iron control 0.004%

Acid 0.123%

Corrosion inhibitor 0.002% Biocide 0.001%

Graph 1 Volumetric percentages of additives in fracturing fluids from Modern Shale Gas Development in the United States.

As described in previous sections, the chemical additives are included in fracking fluids to tailor the fluids to the requirements of the specific geological situation. The very popular chart above taken from Modern Shale Gas Development in the United States demonstrates typical volumetric percentages of additives that were used for a typical hydraulic fracturing treatment of a Fayetteville Shale horizontal well (Graph 1).

3.1.6.1 Proppants Proppants are pretty hard to make into anything fun, exciting or entertaining…as they are, for the most part, made up of sand or a manufactured facsimile of sand. Sure, if you want to be poetic you can refer to proppants as the only materials the operators want to remain downhole in the fractures. If you really want to think poetically, feel free to consider a proppant’s life as one of making its way from origins mined within the Earth only to return to its final resting place deeper within the Earth’s fractures. As discussed earlier, proppants are simply materials (typically silica sand, resin coated silica sand, or manufactured ceramics) used to prop open the open fractures to promote flow and eventual extraction of hydrocarbons. As simple as prop-pants may seem, the estimated amount of proppant used in industry has grown tenfold since 2000. In some regions it is not uncommon to see upwards of four million pounds of proppant used per well and represent up to 5% of well costs. The growth in proppant usage is generally attributed to operators realizing better well completion techniques with more proppant per stage and better well pad techniques with more laterals and fracturing stages per pad.

62

Hydraulic Fracturing and Well Stimulation

Even considering the accelerated growth in the last decade, the evolution of proppant usage has been slow to develop over the industry lifetime as a whole. Consider that the first frac job was conducted in 1947, utilizing a reported approximate 20,000 pounds of uncoated frac sand, and manufactured ceramic proppant was not first used until 1983...or 36 years later. Then, approximately one year later, resin coated proppant was first introduced. As with most all technologies, as new techniques continue to develop, proppants will surely evolve further to increase effectiveness and efficiency in hydraulic fracturing. No matter the type of proppant used, the most important characteristics for a proppant are particle size distribution, crush resistance, shape and sphericity (or roundness). Proppant materials are carefully sorted for size and sphericity to provide an efficient conduit for production of fluid from the reservoir to the well-bore. Grain size is critical because a proppant must reliably fall within certain size ranges to coordinate with downhole conditions and completion design (Figure 3.1). Proppant shape and hardness qualities are also very important to the efficiency and effectiveness of a fracturing operation. A coarser proppant allows for higher flow capacity due to the larger pore spaces between grains, but it may break down or crush more readily under high closure stress, and rounder, smoother proppant shapes allow for better permeability (Figure 3.2). Another important quality that must be taken into consideration is its hardness with respect to the formation. If the proppant is unable to embed in the formation, something referred to as point load occurs, which leads to higher flow capacity, but the proppant will break easier. However, if the proppant is able to embed in the formation, it is referred to as embedment, Sufficiently placed & sized proppant - Effective return -

No proppant - No return -

Individual fracture Return flow Insufficiently placed & sized proppant - Ineffective return -

Figure 3.1 Proppant size and placement. Courtesy of the authors.

Primer on Hydraulic Fracturing 63 Proppant roundness = Effective return and strength

Proppant irregularity = Less effective return and weakness

Figure 3.2 Proppant shape. Courtesy of the authors.

Point load proppant

Higher flow capacity, increased proppant fragility

Embedded proppant

Lower flow capacity, increased proppant strength

Figure 3.3 Proppant hardness. Courtesy of the authors.

which results in the load pressure spreading out over the proppant area, increasing the breaking point but also lowering flow capacity. Embedment is also a function of particle size (Figure 3.3). Even though most all proppant materials are naturally occurring, including manufactured ceramic proppants, with relatively low amounts of additional engineering necessary, the logistics in procuring and transporting proppants can be daunting. Logistical considerations include coordination of manufacturing material resources, transportation costs, and possibly a substantial monetary investment in equipment necessary for processing and material handling facilities.

3.1.6.2 Silica Sand While the all-encompassing term for the material “sand” is generally used for pretty much all forms of broken down granules of minerals or rocks, to be specific falling between silt and gravel in the spectrum of sizes. There are, however, many varieties of sand in the world, each with their own unique composition and qualities. We all like to picture the white sandy beaches of vacation destinations, for example, which are made up primarily of limestone that has been broken down. Then there are also many black

64

Hydraulic Fracturing and Well Stimulation

sands either volcanic in origin or containing magnetite. Other sands have high levels of iron in them, and so are rich and yellow in color. The type of sand utilized for proppant materials is silica sand, which is, by far, also the most commonly used type of proppant. Silica sand, unlike many other “ingredients” of frac fluid, is more of a natural resource than an engineered product. Silica sand proppant is, in a simplistic description, made up of the most common mineral in the Earth’s continental crust… quartz. Silica sand is simply quartz that over the years, through the work of time and several erosion forces, has been broken down into tiny granules. Even though silica sand is a relatively common material, silica sand used for proppant is a specifically selected and utilized product. Proppant quality silica sand is a direct function of both the original depositional environmental and some slight mechanical processing, if necessary. Silica sand used for proppant is chosen for its round spherical shape and commonly graded particle distribution…unlike the common sand you might find at the beach or on a playground, which often feels gritty when rubbed between the fingers. In addition to the oil and gas industry, there is some competition between other industries for the bulk of silica sand, as industrial grade silica sand has a wide range of uses. This resource is also commonly used in the manufacture and preparations of various types of glass, in water filtration, sand blasting, as fill and as an ingredient in industrial concrete, in the metal casting industry to make cores and molds, and ironically it is also used in the creation of highly flame-resistant industrial molds and construction materials for the kilns used in the manufacture of the sintered ceramic and bauxite proppants. Even considering all the helpful and positive uses for silica sand across several different industries, there are some possible hazards related to its use. Because of the fine grains involved in silica sand, it can present a health risk if not properly handled. Care must be taken to keep the silica sand out of the lungs during use, and all materials containing more than 0.1% of silica sand must be clearly labeled. Workplace health applications also need to be in place and enforced – failure to wear a proper respirator or mask can result in lung irritation, and prolonged exposure can cause a chronic condition known as silicosis. Silicosis is a form of lung disease resulting from occupational exposure to silica dust over a period of years – causing a slowly progressive fibrosis of the lungs, impairment of lung function and even a heightened susceptibility to tuberculosis of the lungs. Silicosis can also progress and worsen even after someone is no longer exposed to the silica dust, causing long term effects and shortness of breath years later. Also, in the year 2000, the World

Primer on Hydraulic Fracturing 65 Health Organization determined crystalline silica is “associated with silicosis, lung cancer and pulmonary tuberculosis” in classifying it as a Group I carcinogen “based on sufficient evidence carcinogenicity in humans and experimental animals.”

3.1.6.3 Resin Coated Proppant As the name suggests, and to describe in the most simplistic of terms, resin coated proppant is exactly that – silica sand coated with resin. Resin coating silica sand proppant is utilized for two main functions: 1. to spread the pressure load more uniformly to improve the crush resistance of the silica sand particles 2. to keep pieces together that were broken due to high closure stress from down hole pressure and temperature – this not only prevents broken pieces from flowing into the borehole, but also prevents these same broken pieces from returning to the surface during flowback production operation. Currently there are two types of resin coated proppants, Pre-cured and Curable. Pre-cured is the “original” technology in which the resin coating on the silica sand grains is fully cured prior to injection into the fractures. The newer, curable technology has often been described, and I believe very well described, as having a coating that is not completely “baked” or hardened. Curable resin coated prop-pants are used at a little more than half cure so that when the proppant is pumped downhole it can finish curing in the fractures with down hole pressure and temperature. The advantage to curable proppant technology is that it allows the individual proppant grains to bond together in the fracture – resulting in the grains bonding together uniformly in strength when temperature and pressure reach a appropriate levels.

3.1.6.4 Manufactured Ceramics Proppants A third commonly used type of proppant is the manufactured ceramic proppant. This is a proppant generally manufactured from a type of ceramic material – typically non-metallurgic bauxite or kaolin clay. Bauxite is an aluminum ore from which most aluminum is extracted, while kaolin is one of the most common minerals, occurring in abundance from chemical weathering of rocks in hot, moist climatic soils like tropical rainforest areas. Both bauxite and kaolin are utilized as proppants because of their superior strength characteristics which are further enhanced through a

66

Hydraulic Fracturing and Well Stimulation

process known as sintering. The sintering process is conducted in hightemperature kilns that are used to bake the bauxite or kaolin powder after it has been made into specifically sized particles. This process decreases the water content in the bauxite and kaolin to make them more uniformly shaped for size roundness and spherical shape. The desired results of this process are that the manufactured ceramic proppants can be engineered to withstand high levels of downhole pressure (closure stress).

3.2 Additional Types As more is learned through the ongoing processes and further advances are made in technology, additional types of proppants are sure to come up. One current trend is toward the usage of “waste” material – including glass, metallurgical slags, and even rock cuttings produced to the surface during oil and gas drilling. The re-use of rock cuttings from gas drilling operations is especially attractive, since not only does it re-use a common waste product in industry, but it is also utilizing sources indigenous to the locality – which will cut down wastes while also cutting down on transportation and overhead costs. However, the other possibilities are also quite attractive in that agreements can be made with landfills, metallurgical operations and glass companies to recycle and re-use their wastes in lieu of land filling.

3.3 Other Most Common Objections to Drilling Operations Now you will notice this discussion of common objections to drilling operations is a little different than other objections to drilling presented in this book for two reasons: 1. there is no attempt to separate the operations related to fracking from all other drilling operations – this is simply because most nuisances related to one operation are the same for all (for instance, additional traffic is additional traffic no matter the origination), and 2. the nuisances described in this section are not written of or discussed in a quantifiable way – in other words, this discussion is not centered on an amount, but the simple fact that it exists.

Primer on Hydraulic Fracturing 67 A reason for this is because a lot of data is collected for the other aspects related to fracking operations to try and prove/disprove their existence, when the nuisances discussed in this section are easily seen as in existence (just spend a few minutes along any road used for drilling operations, and this will become abundantly clear). Please keep in mind that there are many additional nuisances absorbed by those living near drilling locations or related roadways, so this listing is far from comprehensive. The following are merely what most see as the most common and are not presented in any order of magnitude. That decision has to be made by each individual – one person may be more affected by noise, while another is much more concerned with dust.

3.3.1 Noise Noise conditions are usually one of the first things to change and be noticed by local landowners. An increase in noise is also one of the most continuous nuisances related to operations. Drilling and completing a well – from the pad construction to the final completion of the well – takes several weeks and utilizes many different types equipment. This additional equipment can include additional trucking, construction and drilling equipment. The noise concerns usually begin with the additional traffic brought to an area during pad construction, then continue with the noises associated with equipment and trucking required to construct a pad, only to be followed by the large amount of noise related to rig construction and operation throughout the well drilling process. Then, once the well site is completed, there may come the additional sounds of compressors used during ongoing production activities. When you think of noise concerns related to the oil and gas industry, the first thing that commonly comes to mind is the big noisy rig or maybe the noisy traffic coming back and forth. These are, of course, very real and valid concerns; however, the thing that is quite possibly the most notable noise nuisance related to the oil and gas industry, due to length of time, is the compressor. For the most part, the heavy rig work and heavy truck traffic lasts approximately one to two months – while the compressor, while not as loud, can continue for a much longer amount of time (months to even years). Gas compressors are normally the largest equipment remaining after the well development process is complete and are utilized for something called gas lift. Gas lift is used in wells that have insufficient reservoir pressure to produce efficiently on their own. The gas lift process involves injecting gas through the tubing-casing annulus to aerate the fluid to reduce

68

Hydraulic Fracturing and Well Stimulation

its density. Following aeration of the fluid the formation pressure is then able to better lift the oil column up the wellbore. For pad sites where long term compressor use is anticipated, especially in rural communities where serenity is the norm and even the slightest ongoing noise can be heard clearly for long distances, operators have addressed compressor noise concerns with remote siting (trying to locate the compressors on the part of the pad farthest from homes), noise tampering sound walls, and directing compressors with fans away from homes. However, even with the measures presently taken to mitigate ongoing sound issues, additional work must be done and technology developed to work toward a solution.

3.4 Changes in Landscape and Beauty of Surroundings Several different types of pollution are commonly mentioned in relation to the oil and gas industry – including water pollution, soil pollution, air pollution and, as presented in the previous section, noise pollution. However, one that may be overlooked to the majority of the public, but certainly not overlooked to those affected, is visual pollution. Visual pollution is an aesthetic issue, “referring to the impacts of pollution that impair one’s ability to enjoy a vista or view.” Now, with the possible exception of the immense number of billboards lining our nation’s highways, not many things meet the definition of visual pollution as much as a drill rig. Drill rigs utilized in most unconventional well drilling typically can range from approximately 50 feet to 100 feet in height. Couple the height of the drill rig with the ongoing movement and dust related to drilling, and it is easy to imagine how this would be bothersome to those adjacent to rig locations. One mitigation attempt for this problem would be the usage of lower height rigs. However, the undesirable trade-off for a lower height rig is the necessary extended time on location for smaller rigs. Ironically, horizontal drilling techniques commonly related to unconventional well drilling and hydraulic fracturing locations can actually be considered a “semi-solution” to this problem. Pads used for horizontal drilling commonly include multiple laterals on one location, in which the drilling of multiple wells literally means moving the rig over as little as twenty feet from one location on the pad to the next. This allows wells to be drilled from one location without the necessity of moving the rig and drilling in several locations – which would only disturb that many more possible visual pollution points. This also allows for accelerated drilling time due to lessened rig movement time, a reduction in the number of

Primer on Hydraulic Fracturing 69 necessary lease roads and drill pad locations, fewer necessary pipelines and fewer tank batteries.

3.5 Increased Traffic Another nuisance commonly cited by those living in oil production areas is the drastic amount of added traffic it creates. This is not necessarily the type of traffic most of us think of when we hear the word. Traffic related to the oilfield includes all the initial traffic to bring in heavy equipment for pad construction and eventually the rig itself, followed by traffic for well completion and fracking activities (to get a taste for what this is like, consider the amount of sand used in each frac job then consider how many separate truckloads that would be), then the ongoing traffic related to hauling produced water and oil from the locations until some sort of pipeline infrastructure can be put in place. Also, keep in mind that many of the areas affected by oil and gas operations are rural and do not, quite simply put, have the proper roadways for the larger size or amount of traffic vehicles that come with industry operations. Not only does the added traffic add additional wear and tear to the local roadways, the narrower two-lane, and sometimes even more narrow gravel roads, cause very unsafe driving conditions for the industry and local resident vehicles alike. The answers to the traffic problems may seem obvious – do something to lessen the amount of traffic or do something to improve the roads – but finding ways to turn those answers into reality is something much more difficult than may first appear. The first, “do something to lessen the amount of traffic,” would include: 1. the need to either use fewer (but larger) transport vehicles – resulting in additional hazardous conditions with the larger vehicles on the narrow rural roads, or 2. the need to install a pipeline infrastructure to transport produced water and/or oil – which comes with the obvious concerns related to pipeline installation and location. The second, “do something to improve the roads,” would depend on the type of road to be improved. Improving and widening gravel type lease and rural roads is a less daunting task than improving paved city/county roads due to ease of obtaining the proper materials and the fewer restrictions put on maintenance. However, making improvement to city and county roads

70

Hydraulic Fracturing and Well Stimulation

would include needing to clear a wider right of way all along the road to be widened and the city/county would need to have the funds set aside for this task…which is a time consuming process.

3.6 Chemicals and Products on Locations Another common point of contention for residents in oil producing areas is on-site storage of chemicals and products. For decades, one of the biggest drivers for public concern has been the identity and amounts of chemicals stored on pad locations during all phases of the well completion and production process. This can include fuels used on location, the makeup of drilling fluids (water based, oil based or synthetic), fracking chemicals and additives stored on locations, chemicals kept on-site during ongoing production, and even the products – produced water and/or oil – stored on site as recovered from the well during operations. The general public would be pretty surprised at how little chemical and products are stored on site during construction, drilling and frack operations. For the most part, oilfield operations have become such a streamlined and efficient operation that operators will know how much of a given chemical product will be necessary and, for the most part, make all attempts to have the chemicals arrive on location as close to when needed as possible to avoid storage. This process is beneficial to the operator in that it cuts down on the time taken up by storing chemicals only to return for them when needed, cuts down on waste from unused or outdated chemicals, cuts down on equipment needed to maneuver chemicals if you have them delivered directly to point of need, allows for more working space on the pad, and also helps avoid a great deal of logistical problems related to maneuvering equipment around storage areas. Once production operations are in place and wells begin producing, the fluids – produced water and oil – are often stored on site in large tanks while awaiting transport off-site. Safeguards put in place to protect the environment and public from tank releases include consistent measurements by pumpers, high level shut down sensors, continued equipment observations and maintenance, and secondary containments in place around the tanks to contain any fluids that may release. Secondary containments may be constructed of properly packed and integrity tested earthen materials or up to specifically designed and manufactured metal containments with plastic liners. No matter the materials used in construction, secondary containments must be sufficiently large enough to contain all the fluids that

Primer on Hydraulic Fracturing 71 could possibly escape the tanks plus sufficient extra space for “worst case scenario” rainfall. This amount is calculated for each region of the country based on historic rainfall data. Even with attempts to minimize the amount of on-site storage, some chemical and product storage is unavoidable, and there are very valid concerns, including potential spills, leaks, tank or container overfill, and even the chance of traffic accidents on location or roadways leading to releases of chemicals and/or products. Release events could range from relatively small amounts from equipment leaks to possibly hundreds of barrels from tank release. Two regulatory measures in place to manage and oversee on-site chemical storage conditions are requiring Spill Prevention Countermeasure and Control (SPCC) plans and SARA reporting. SPCC plans are documents required by all facilities having the potential to discharge oil to navigable waters of the U.S. and meeting one or both of the following: greater than 1,320 gallons (31.4 bbls) aggregate aboveground storage in equipment, drums, tanks, totes, tanks greater than 55 gallons in size; or, greater than 42,000 gallons total underground storage capacity. Just to clarify, aggregate refers to adding up separate amounts of all storage vessels…you can have one 1,320 gallon tank or ten 132 gallon tanks and they would be equal under the SPCC requirements. Also, the “having potential to discharge oil to navigable waters of the U.S.” is left up to regulatory discretion to calculate, and has come to include pretty much anywhere in the U.S. you could imagine. SPCC plans are, to keep it simple, engineer-stamped documents that must be created for all facilities meeting the above conditions that include a list of spill response procedures, an emergency notification phone list, inspection procedures and schedule, training requirements, site figures, site chemical and product storage vessel types and sizes and containment calculations to prove sufficient containment is given to contain the largest possible spill amount. SARA reporting, or possibly better known as the federal “right to know,” requires quarterly and annual reporting of chemical storage details (types of chemicals, amounts and dates of storage) for all facilities which used more than 10,000 pounds per year of the chemical exceeding the threshold quantity. This requirement means a facility storing more than 10,000 pounds of a given chemical in a year must report that chemical and amount. This program is intended as the “right to know” for emergency responders and emergency services that may respond to an emergency situation on the location so they will be able to adequately prepare for what may be stored on site. The drawback of this program as related to the oil and gas industry is that, with quarterly reporting,

72

Hydraulic Fracturing and Well Stimulation

by the time a chemical has been reported the oilfield function requiring the chemical has normally been long complete and the chemicals are no longer on site. This basically means that once the chemical is reported as being on a location it is no longer there; however, as previously stated, oilfield operations have become such a streamlined process that if you know what has been reported for a previous location by a specified operator you can, for the most part, expect much the same chemicals and products stored at following locations. If you are really curious about all the chemicals used at a site, ask to receive a copy of the Material Safety Data Sheet of the chemicals used.

3.6.1 Material Safety Data Sheets (MSDS) Anytime a company produces for sale or uses a chemical a Material Safety Data Sheet (MSDS) has to be written on the product and on file when used. Occupational Safety and Health Administration (OSHA) estimates that there are over 650,000 hazardous chemicals used daily in the United States, and that hundreds more will be added this year alone. To address the physical and health hazards of these chemicals, OSHA finalized the Hazard Communication Standard (HCS) on November 25, 1983. The purpose of the HCS is to “ensure that the hazards of all chemicals produced or imported are evaluated, and that information concerning their hazards is transmitted to employers and employees.” (29 CFR 1910.1200(a)(1)). Employers are under obligation to use labels, MSDS, and other information to evaluate both the physical and health hazards created by the use of chemicals in their workplace, establish a program that addresses these hazards and train workers to minimize their exposure. According to an OSHA Executive Summary, “Chemical information is the foundation of workplace chemical safety programs. Without it, sound management of chemicals cannot occur. The HCS has made provision of hazard information about chemical products an accepted business practice in the United States. There is now a whole generation of employers and employees who have never worked in a situation where information about the chemicals in their workplace is not available.” Manufacturers or importers of chemicals must create or obtain a MSDS for every hazardous chemical that they produce or import (29 CFR 1910.1200(g)), and supply the appropriate one with a customer’s first purchase, and any time the MSDS changes. (29 CFR 1910.1200(g)(6)(i)). Employers are not required to evaluate information on a MSDS. (29 CFR 1910.1200(d)(1)). They do, however, have a duty to study and to use it to

Primer on Hydraulic Fracturing 73 “develop, implement and maintain... a written hazard communication program” to ensure the safety of their workers (29 CFR 1910.1200(e) (1)). To help address worker safety “at all times,” OSHA requires employers to make MSDS “readily accessible during each work shift to employees when they are in their work area(s).” (29 CFR 1910.1200(g)(8)). OSHA permits electronic and other forms of access to MSDS, as long as there are “no barriers to immediate employee access in each workplace.” (29 CFR 1910.1200(g)(8)).

3.6.1.1 Contents of an MSDS Frac site workers as well as anyone working in an industry or market that uses chemicals will have access to an MSDS for any chemical that they may have contact with. Interestingly enough, consumer products also have MSDSs. In fact, your local hardware store has a complete file of MSDSs for all the consumer chemicals they sell and many department stores do as well. If you ever want to know the dangerous effects of a particular insecticide or cleaner you can refer to the MSDS for detailed information. Beware though; the information contained within an MSDS can be a bit foreboding. Like pharmaceuticals and over-the-counter medications, the warnings typically are meant to take into consideration any and all dangers that may happen upon exposure. Without a working knowledge of the terms and criteria put forth in the MSDS, the layperson could quickly become horrified with the prospect of using a product only to experience dizziness, dry mouth or shortness of breath (which seems to be the universal response to everything from aspirin to Zoloft ). The following is an explanation which is provided to help you interpret the information found on manufacturers’ MSDSs. While the format of these data sheets varies from manufacturer to manufacturer, certain components appear on each sheet.

3.6.1.2 Product Identification The MSDS shall provide the name and address of the manufacturer and an emergency phone number where questions about toxicity and chemical hazards can be directed. Product Name: Commercial or marketing name. Synonym: Approved chemical name and/or synonyms. Chemical Family: Group of chemicals with related physical and chemical properties.

74

Hydraulic Fracturing and Well Stimulation Formula: Chemical formula, if applicable; i.e., the conventional scientific definition for a material. CAS Number: Number assigned to chemicals or materials by the Chemical Abstracts Service.

3.6.1.3 Hazardous Ingredients of Mixtures The MSDS shall describe the percent composition of the substance, listing chemicals present in the mixture. If it was tested as a mixture, it lists chemicals which contribute to its hazardous nature. Otherwise, it lists ingredients making up more than 1% and all carcinogens. The OSHA permissible exposure limit (PEL), National Institute for Occupational Safety and Health (NIOSH) recommended exposure limit (REL), and/or the American Conference of Governmental Industrial Hygienists (ACGIH) threshold limit value (TLV) will also be listed, if appropriate. The OSHA PEL is the regulated standard, while the others are recommended limits. The PEL is usually expressed in parts per million parts of air (ppm) or milligrams of dust or vapor per cubic meter of air (mg/m3). It is usually a time weighted average (TWA) – concentration averaged over an eight hour day. Sometimes, a short term exposure limit (STEL) may be listed. The STEL is a 15 minute TWA which should not be exceeded. A ceiling limit, (c), is a concentration which may not be exceeded at any time. A skin notation means that skin exposure is significant in contributing to the overall exposure.

3.6.1.4 Physical Data The MSDS shall outline the physical properties of the material. The information may be used to determine conditions for exposure. For example, one can determine whether or not a chemical will form a vapor (vapor pressure), whether this vapor will rise or fall (vapor density), and what the vapor should smell like (appearance and odor). This could help determine whether to use a fume hood or where to place ventilators. The following information is usually included: Boiling Point: temperature at which liquid changes to vapor state. Melting Point: temperature at which a solid begins to change to liquid. Vapor Pressure: a measure of how volatile a substance is and how quickly it evaporates. For comparison, the VP of water

Primer on Hydraulic Fracturing 75 (at 20°C) is 17.5 mm Hg, Vaseline (non-volatile) is close to 0 mm Hg, and diethyl ether (very volatile) is 440 mm Hg. Vapor Density (air = 1): weight of a gas or vapor compared to weight of an equal volume of air. Density greater than 1 indicates it is heavier than air, less than 1 indicates it is lighter than air. Vapors heavier than air can flow along just above ground, where they may pose a fire or explosion hazard. Specific Gravity (water = 1): ratio of volume weight of material to equal volume weight of water. Solubility in Water: percentage of material that will dissolve in water, usually at ambient temperature. Since the much of the human body is made of water, water soluble substances more readily absorb and distribute. Appearance/Odor: color, physical state at room temperature, size of particles, consistency, odor, as compared to common substances. Odor threshold refers to the concentration required in the air before vapors are detected or recognized. % Volatile by Volume: Percentage of a liquid or solid, by volume, that evaporates at a temperature of 70°F. Evaporation Rate: usually expressed as a time ratio with ethyl ether = 1, unless otherwise specified. Viscosity: internal resistance to flow exhibited by a fluid, normally measured in centistokes time or Saybolt Universal Secs. Other Pertinent Physical Data: information such as freezing point is given, as appropriate.

3.6.1.5 Fire & Explosion Hazard Data The MSDS shall include information regarding the flammability of the material and information for fighting fires involving the material. Flashpoint: the lowest temperature at which a liquid gives off enough vapor to ignite when a source of ignition is present. Auto-ignition Temperature: the approximate temperature at which a flammable gas-air mixture will ignite without spark or flame. Vapors and gases will spontaneously ignite at lower temperatures in oxygen than in air. Flammable Limits: the lower explosive limit (LEL) and upper explosive limit (UEL) define the range of concentration of a gas or vapor in air at which combustion can occur. For

76

Hydraulic Fracturing and Well Stimulation instance, an automobile carburetor controls this mixture – too lean (not enough chemical) or too rich (not enough air, as when you flood your engine) will not ignite. Extinguishing Media: appropriate extinguishing agent(s) for the material. Fire-fighting Procedures: Appropriate equipment and methods are indicated for limiting hazards encountered in fire situations. Fire or Explosion Hazards: Hazards and/or conditions which may cause fire or explosions are defined.

3.6.1.6 Health Hazard Data The MSDS shall define the medical signs and symptoms that may be encountered with normal exposure or overexposure to this material or its components. Information on the toxicity of the substance may also be presented. Results of animal studies are most often given, i.e., LD50 (mouse) = 250 mg/kg. Usually expressed in weight of chemical per kg of body weight. LD50 or lethal dose 50 is the dose of a substance which will cause the death of half the experimental animals. LC50 is the concentration of the substance in air which will cause the death of half the experimental animals. Health hazard information may also distinguish the effects of acute (shortterm) and chronic (long-term) exposure.

3.6.1.7 Reactivity Data The MSDS shall include information regarding the stability of the material and any special storage or use considerations. Stability: “unstable” indicates that a chemical may decompose spontaneously under normal temperatures, pressures, and mechanical shocks. Rapid decomposition produces heat and may cause fire or explosion. Conditions to avoid are listed in this section. Incompatibility: certain chemicals, when mixed may create hazardous conditions. Incompatible chemicals should not be stored together. Hazardous Decomposition Products: chemical substances which may be created when the chemical decomposes or burns. Hazardous Polymerization: rapid polymerization may produce enough heat to cause containers to explode. Conditions to avoid are listed in this section.

Primer on Hydraulic Fracturing 77

3.6.1.8 Personal Protection Information The MSDS shall include general information about appropriate personal protective equipment for handling this material. Many times, this section of the MSDS is written for large scale use of the material. Appropriate personal protection may be determined by considering the amount of the material being used and the actual manipulations to be performed. Eye Protection: recommendations are dependent upon the irritancy, corrosiveness, and special handling procedures. Skin Protection: describes the particular types of protective garments and appropriate glove materials to provide personnel protection. Respiratory Protection: appropriate respirators for conditions exceeding the recommended occupational exposure limits. Ventilation: air flow schemes (general, local) are listed to limit hazardous

3.7 Conclusion One hundred and fifty years ago when crude oil was first being extracted, the damage done to the environment was nothing short of a nightmare. In some areas little has changed but in many instances companies take extraordinary precautions. Many years ago, we did not understand the ramifications of pollution. Today, much work is underway to address what is being understood as environmental concerns. The work that goes into preparing a fracing well site today in assuring that the chemicals used are innocuous while maintaining the physical integrity of the surrounding land is taken into consideration primarily for legal reasons as well as business concerns. If environmental laws are in place, then work shall be structured accordingly. The challenge is to enact law that makes sense according to empirical evidence. It is also vital that those using the process and chemicals as well as the public understand the technology. Where industry in general fails is when it rushes forward and only applies the letter of the law and does not push for higher requirements. Legislation is often left to the most verbal and most passionate. It is fair to say that when emotion runs high, logic wanes. If we are to utilize the gifts that the Earth has bestowed upon us then it is more than fair to assume a protective role going forward and exploring proper protocols to ensure that the environment and comfort of life remain as balanced as it was found.

78

Hydraulic Fracturing and Well Stimulation

Bibliography Fjaer, E. (2008). “Mechanics of hydraulic fracturing”. Petroleum related rock mechanics. Developments in petroleum science (2 ed.). Elsevier. p. 369. ISBN 978-0-444-50260-5. http://books.google.co.uk/books?id=l6CfasFxhzYC&pg= PA369&lpg=PA369. Laubach, S.E.; Reed R.M., Olson J.E., Lander R.H. & Bonnell L.M. (2004). Coevolution of crack-seal texture and fracture porosity in sedimentary rocks: cathodoluminescence observations of regional fractures“. Journal of Structural Geology (Elsevier) 26 (5): 967–982. doi:10.1016/j.jsg.2003.08.019. http://www.sciencedirect.com/science/article/pii/S0191814103001858. “Definition of frac job”. Oilfield Glossary. Schlumberger. http://www.glossary.oilfield.slb. com/Display.cfm?Term=frac%20job. Office of Air and Radiation U.S. Environmental Protection Agency (November 2010). “General Technical Support Document for Injection and Geologic Sequestration of Carbon Dioxide: Subparts RR and UU Greenhouse Gas Reporting Program”. Office of Air and Radiation U.S. Environmental Protection Agency. http://www.epa.gov/cli-matechange/emissions/downloads10/Subpart-RR-UU_TSD.pdf. Jack E. Whitten, Steven R. Courtemanche, Andrea R. Jones, Richard E. Penrod, and David B. Fogl (Division of Industrial and Medical Nuclear Safety, Office of Nuclear Material Safety and Safeguards (June 2000). “Consolidated Guidance About Materials Licenses: Program-Specific Guidance About Well Logging, Tracer, and Field Flood Study Licenses (NUREG-1556, Volume 14)”. US Nuclear Regulatory Commission. http://www.nrc. gov/reading-rm/ doc-collections/nuregs/staff/sr1556/v14/#_1_26. Anthony Andrews et al. (30 October 2009) (PDF). Unconventional Gas Shales: Development, Technology, and Policy Issues. Congressional Research Service. p. ?. http://www.fas.org/sgp/crs/misc/R40894.pdf. Bill McKibben (8 March 2012). “Why Not Frack?”. The New York Review of Books 59 (4). http://www.nybooks.com/articles/archives/2012/mar/08/why-not-frack/. David Biello (30 March 2010). “What the Frack? Natural Gas from Subterranean Shale Promises U.S. Energy Independence--With Environmental Costs”. Scientific American. http://www.scientificamerican.com/article. cfm?id=shale-gas-and-hydraulic-fracturing. (PDF) Chemicals Used in Hydraulic Fracturing (Report). Committee on Energy and Commerce U.S. House of Representatives. April 18, 2011. http:// democrats.energy-commerce.house.gov/sites/default/files/documents/ Hydraulic%20Fracturing%20 Report%204.18.11.pdf

4 A Graph Theoretic Approach for Spatial Analysis of Induced Fracture Networks Deborah Glosser1,2* and Jennifer R. Bauer3,4 1

United States Department of Energy, National Energy Technology Laboratory, Pittsburgh, PA, USA 2 Oak Ridge Institute for Science Education, Oak Ridge, TN 3 United States Department of Energy, National Energy Technology Laboratory Albany, Oregon, USA 4 United States Department of Energy, National Energy Technology Laboratory, AECOM, Albany, Oregon, USA

Abstract Drilling induced fractures are generated when excessive stresses around a borehole cause tensile failure of the wellbore wall. If stress concentrations are great enough, compressive failures can form in the region surrounding the wellbore, leading to wellbore breakout, and the potential compromise of wellbore integrity. Another category of induced fracture networks are hydraulically induced fractures, which are generated by the injection of pressurized fluids into the subsurface. Overlapping induced fracture networks between collocated wellbores may increase pathways in the subsurface, and create the potential for unwanted fluid leakage. The generation of induced fractures is greatly dependent upon the structural and geological characteristics. Probabilistic-based simulations are often used to model fracture systems. Several methods for modeling local fracture networks have been proposed in the literature. These models often involve the generation of randomly located fractures, and may have limited capabilities for honoring engineered fractures such as induced fracture networks. We present a graph theoretic approach for identifying geospatial regions and wellbores at increased risk for subsurface connectivity based on wellbore proximity and local lithologic characteristics. The algorithm is coded in Matlab, and transforms 3 dimensional geospatial data to graph form for rapid computation of pairwise and topological relationships between wellbores (nodes), and the spatial radius of induced fractures (edges). *Corresponding author: [email protected] Fred Aminzadeh (ed.) Hydraulic Fracturing and Well Stimulation, (79–97) © 2019 Scrivener Publishing LLC

79

80

Hydraulic Fracturing and Well Stimulation

Induced fracture reaches are represented as cylinders with a radius r, based on literature derived ranges for fracture lengths for different lithologies (e.g. shale, sandstone). The topological algorithm is compared to a standard graph-based k-nearest neighbor algorithm to demonstrate the value of incorporating lithologic attributes in graph-based fracture models. The algorithms are applied to two scenarios using Pennsylvania wellbore and lithologic data: a subset of data from the Bradford field, as well as a known leakage scenario in Armstrong County. The topological algorithm presented in this paper can be used to complement existing fracture models to better account for the reach of induced fractures, and to identify spatial extents at increased risk for unwanted subsurface connectivity. As a result, the method presented in this paper can be part of a cumulative strategy to reduce uncertainty inherent to combined geologic and engineered systems. The model output provides valuable information for industry to develop environmentally safe drilling and injection plans; and for regulators to identify specific wellbores at greater risk for leakage, and to develop targeted, science-based monitoring policies for higher risk regions. Keywords: Graph theory, spatial analysis, hydraulic fracturing

4.1 Background and Rationale Commercial hydrocarbon drilling began as early as the 1800s in West Virginia and Pennsylvania [1]. Since then, human engineering of the subsurface has expanded worldwide to include gas storage [2]; CO2 injection [3]; unconventional resource exploration [4]; and injection of hazardous waste [5]. As the character and degree of subsurface activities has expanded, it has become increasingly important to develop methods and techniques capable of addressing the interactions between engineered features and local geology. Information provided by such techniques is of critical value to both industry and regulators: leakage via wellbores and fracture networks is a documented concern [6], and the data provided by such methods is imperative for the development of environmentally safe drilling and injection plans, as well as science-based monitoring and plugging plans for regions (or particular wellbores) at greater risk of connectivity and leakage. Engineered and induced fractures, such as drilling induced fractures (DIFs) and hydraulically induced fractures (HIFs), are important phenomena that can occur when a wellbore is drilled. DIFs are generated when stresses around a borehole are in excess of those required to cause tensile failure of the wellbore wall [7]. If stress concentrations are great enough, compressive failures can form in the region surrounding the wellbore, leading to wellbore breakout, and the potential loss of wellbore

Spatial Analysis of Induced Fracture Networks 81 integrity [8]. HIFs are pressure induced fractures that are generated when fluid is injected at high pressure into subsurface formations. If induced fracture networks around collocated wellbores intersect, there may be an increased likelihood of communication between wellbores, and the potential for unwanted fluid leakage between such networks. This is particularly of concern for older, structurally unstable wellbores, which may lack adequate casing or cementing necessary for zonal isolation [9]. Furthermore, wellbore spatial densities are likely to be relatively high in regions with historical drilling activity: Before regulations requiring minimum wellbore spacing were implemented, it was a common practice to drill multiple wellbores in close proximity (Figure 4.1). Furthermore, regions with extensive drilling histories likely have higher wellbore densities as a result of the multiple centuries of exploration of resources in those regions [1]. The generation of induced fractures is greatly dependent upon the structural and lithological characteristics of local geology, which is often difficult to accurately characterize in the absence of expensive geophysical surveys. Consequently, probabilistic-based simulations are often used to model such fracture systems. Several methods for modeling local fracture networks have been proposed in the literature [10–12]. These models often involve the generation of randomly located fractures, with varying degrees of user defined connectivity controls. Because of the importance of wellbore locations; spatial densities; and the potential for overlapping induced fracture networks to create fluid flow pathways, it is important to

Figure 4.1 Historical photograph of spatially dense wellbores in TItusville, PA (http:// www. acceity.org/2010/09/oil-in-them-thare-hills/).

82

Hydraulic Fracturing and Well Stimulation

account for the probable radius of influence of induced fractures around a wellbore, as determined by the local geologic and geospatial attributes. A model capable of incorporating these factors would complement advanced fracture and fracture flow modeling methods. The data provided by such an approach would allow for improved identification of spatial areas at higher risk for communication between wellbores and geologic networks, and allow for the identification of specific regions and wellbores at increased risk for unwanted leakage. Additionally, such data would provide critical information needed for finer scale study or simulation of these spatial extents and wellbores. The data provided by such an approach supports a range of cumulative risk reduction and modeling strategies due to improved uncertainty constrains regarding subsurface engineeredgeologic system characteristics. The information provided by this data is of critical importance to both industry and regulators. Due diligence for wellbore construction and injection plans – as well as monitoring and plugging of wellbores – requires data on the relative likelihood of leakage potential in particular areas and through specific wellbores. The model presented in this paper supports these uses. Because fracture and fracture flow models tend to cross into the realm of big data, computational efficiency is an important consideration for the development of an integrated method for modeling the relationship between geologic and engineered systems. Traditional geospatial models often require access to financially costly methods such as a geographic information system (GIS). Such traditional geospatial models tend to be either cell or raster based, and hence potentially computationally expensive when applied to big data sets. The developing field of graph theory allows for the representation, storage, and manipulation of geospatial data in the form of graphs, which can provide improved computational efficiency when certain graph theoretic data structures are employed [13]. Broadly speaking, graph theory is a field of mathematics and computer science which involves the study of graphs [14]. Graphs are mathematical structures that can represent real world data, and may be used to model pairwise relationships between objects. A graph structure consists of vertices (sometimes called “nodes”) and edges. Or, stated more rigorously, a graph G, consists of two discrete sets, V (vertices) and E (edges). The present work specifically uses what is called an “undirected graph,” in which the elements of E are unordered pairs of vertices. The vertex set of a graph G is denoted by V(G), and the edge set as E(G) [13]. Graphs are naturally suited for visual representation of spatial relationships: Although the graph structure itself can be either list or matrix based, graph diagrams – such as the example shown in Figure 4.2 – show how elements in a graph lend

Spatial Analysis of Induced Fracture Networks 83

Figure 4.2 A drawing of an example graph structure showing vertices (blue circles) and edges (orange lines).

themselves to visualization. The vertices in a graph can represent almost any type of data (both abstract and discrete), and likewise, the edges can represent multiple types of relationships between the vertices. Edges of a graph are often associated with a weight function, w(e), that maps each edge e in E to a number. The types of relationships represented by the edges and their associated weights can be as simple as Euclidean distance [14], or as complex as spatio-temporal interactions in complex networks [15, 16]. The graph-based “topological” model presented in this paper (coded in Matlab) employs such graph structures in a novel method for characterizing the spatial radius of influence of induced fractures around a wellbore, and the spatial extents potentially at greater risk for unwanted fluid migration. The algorithm is further compared and contrasted to a standard graph-based k-nearest neighbor algorithm, to demonstrate the importance of incorporating lithologic factors into induced fracture and wellbore connectivity models. The topological model can be used to complement existing fracture models to better account for the reach of induced fractures around a wellbore, and to identify potentially connected wellbores and spatial extents for additional investigation as part of a cumulative strategy to reduce uncertainty inherent to combined engineered and geologic systems.

4.2 Graph-Based Spatial Analysis The general workflow for the model is shown in Figure 4.3, and referenced and described in detail in the forthcoming sections. The model algorithm is coded in Matlab and is developed for easy integration with other commercially available software.

84

Hydraulic Fracturing and Well Stimulation

Acquire data Define input parameters Create vertices Define topologic parameters

Create edges kNN approach Generate fracture propogation volume

Create edges topologic approach

Return graph analysis

Figure 4.3 Workflow for preforming a graph analysis to spatially assess induced fractures.

4.2.1 Acquire Geologic Data and Define Regional Bounding Lithology Subsurface geology – particularly deeper lithology and structure – is impossible to accurately characterize. Even with access to expensive geophysical surveys or map databases, it is often the case that only limited information on the rock type or in situ structural characteristics of strata at certain depth intervals is known [17]. However, regional-scale geological databases (maps, surveys, core logs, stratigraphic columns, or well logs) representing depth-dependent lithology and stresses are generally freely available from state agencies [18] and other public sources. The information provided by such resources can provide sufficient information from which to determine the most geologically brittle and/or overstressed lithology in a given region. Such a lithology can be considered to be the “bounding lithology”, or the regional geologic media that provides the physical bounding conditions for the maximum fracture radius in that region. Arguably, these freely available data sources provide the best information for determining the bounding lithology, given the inherent ambiguities in interpretation of geophysical surveys (such as gravity or magnetic anomalies). Furthermore, given the technological limitations of resolving small scale fractures in geologic media (as well as the changes to fracture networks caused by drilling), finer scale data sources would not necessarily yield much more useful information regarding the present state of the geology. Once the bounding lithology is identified from the map surveys, a range for the probable fracture radii can be assigned based off existing literature (Table 4.1). One of the key features of the topological model described in this paper is the adaptability of the analysis for user defined needs: Allowing the user to select the range of fracture radii based on the bounding lithology

Spatial Analysis of Induced Fracture Networks 85 Table 4.1 Literature derived values for the average fracture radii for a bounding lithology type [17, 31–34].

Induced fracture radius (m)

Shale

Sand

67

113

and associated values reported in Table 4.1 (or other user defined values), gives the model flexibility for a range of geologic conditions.

4.2.2 Details of the Topological Algorithm The topological algorithm developed in this model is based on the principle of cylindrical intersection. Wellbore point data are treated as graph nodes, and are imported by the user. The wellbore points (x, y, z) are converted to graph structure by the program. This is accomplished by transforming the geographic coordinates of the wellbore data to an orthogonal, earth fixed frame of reference and Cartesian coordinate system, so that curvature effects can be simplified, and Euclidean distances can be calculated in common units such as meters [19]. Each wellbore node has an associated radius of influence (r), based on the literature derived bounding lithology values (as described in the previous section). The bounding lithology for each wellbore is represented as a numeral in the Matlab program (1 for shale, 2 for sandstone, etc), and the data array for each wellbore node is tagged with its bounding lithology value (alternatively, this numeral value can be entered manually for each wellbore within Matlab, or it can be entered as a user defined radius length value within the program). The bounding lithology represents the most fracture prone strata that the wellbore penetrates, and hence approximates the physical boundary of the potential fracture reach of a wellbore intersecting that strata. Computationally, this is represented as a cylinder’s radius. The model computes the associated radius of influence around each wellbore node in 3-dimensions. This results in a series of finite cylinders of radius r, associated with each graph node. The algorithm computes the intersection of the cylinders on the graph, and when such intersections are identified, an edge is drawn between the originating nodes. This edge indicates that these originating wellbore nodes fall within the potential fracture reach zone of each other.

4.2.2.1 Data Acquisition, Conditioning and Quanta The non-zero graph entries for the topological algorithm are stored in the form of a sparse matrix, making storage and manipulation of large data

86

Hydraulic Fracturing and Well Stimulation

quanta computationally feasible via compressed sparse row indexing. However, certain practical data acquisition and conditioning procedures are necessary to apply the algorithm to the geologic and geospatial data that the algorithm is designed for. As described in prior sections, it is necessary for the user to have a priori knowledge of the predominant geologic characteristics in the subsurface. These data should be derived from geologic maps, surveys, or well logs. In certain cases – particularly where a user is running the algorithm over large spatial extents with multiple wellbores – it may be necessary for the user to perform an initial GIS-based analysis overlaying wellbore data with geologic layers to determine the bounding lithology. The bounding lithology of each wellbore must then be converted to a numeric representation (e.g. “1” for shale, “2” for sandstone), and stored as an element in the row associated with the wellbore points, before it is imported for analysis into the Matlab program. The topological algorithm is also capable of allowing the user to define by hand, both the length of the radius of influence around each wellbore, or simply select by hand the bounding lithology from the pre-populated values within Matlab, when running the algorithm for smaller datasets.

4.2.2.2 Details of the k-Nearest Neighbor Algorithm Many traditional geospatial models rely on nearest-neighbor associations to assess spatial relationships between geospatial features. For this reason, a standard k-nearest neighbor algorithm is presented and applied to the same data as the topological algorithm, as a means to demonstrate the value added by the latter. Details of the knn algorithm can be found in multiple literature sources [20–22]. The basic principle behind the knn approach is that the data points (here, well-bores) exist in a metric feature space. The algorithm is only distance based, and does not integrate any information on lithology or wellbore spatial densities. The user must define the k, or number of nearest neighbors around each wellbore, and the algorithm makes a distance-based selection.

4.2.3 The Value of the Topological Approach Algorithm To demonstrate the value added by the newly developed topological algorithm, it is compared and contrasted to the standard knn algorithm. The knn algorithm establishes an edge between each node and its k closest neighbors (in Euclidean distance) for some user-specified integer k. The topological approach – which is the primary output of the method – considers both geologic and topologic relationships by computing edges based on

Spatial Analysis of Induced Fracture Networks 87 bounding lithologies and associated radii of influence. Comparing and contrasting the outputs from two algorithms serves to highlight the importance of accounting for subsurface geologic features in representing real world geospatial data in graph form. A real world example is demonstrated in the sections below.

4.3 Real World Applications of the Algorithm 4.3.1 Bradford Field: Contrasting the Graph-Based Approaches; k Sensitivity Commercial hydrocarbon exploration has been occurring in the Bradford field in Pennsylvania since the 1800s [9]. The extensive drilling history in the Bradford field results in a high spatial density of wellbores; particularly older “legacy” well-bores which are likely to be structurally unsound and poorly sealed [1] (Figure 4.4). The model is demonstrated using a synthetic subset of spatially continuous well-bore data in this region.

Bradford oil & Gas field Pennsylvania wellbores Status of wellbores in bradford eld Active Plugged &/or abandoned Other N

0

1 .5

3

4. 5

6

Kilometers

Figure 4.4 Distribution and known status of wellbores within the Bradford Oil & Gas field, Pennsylvania.

88

Hydraulic Fracturing and Well Stimulation

4.3.1.1 Data Sources The Bradford PA wellbore data used in this analysis were obtained from the National Energy Technology Laboratory, United States Department of Energy (NETL/DOE), and were part of an aggregated dataset produced using the methodology described by Dilmore et al. [9] and Glosser et al. [1]. Initial data provenance for the dataset includes, aeromagnetic surveys [23], digital databases [24, 25], and historical maps and minerals reports [26, 27]. Bounding lithologies were determined either by the individual wellbore records (where available), or by performing a geospatial overlay of the wellbore locations with a freely available geologic map [28].

4.3.1.2 Results The two algorithms – the knn algorithm and the topological algorithm, were executed on these data. 50 wellbore locations from the Bradford field were subsampled, and associated bounding lithology values for the wellbores were chosen. The knn algorithm was run for three nearest neighbor scenarios: k = 1 (Figure 4.5f); k = 2 (Figure 4.5e); and k = 3 (Figure 4.5d). For the topological algorithm, the results are presented in 3 dimensions with the induced fracture radius of influence for each wellbore (Figure 4.5a); in 2D form with the radius of influence (Figure 4.5b); and in simple graph form showing only the edge connections (Figure 4.5c). Unlike the knn approach – where edges are drawn to each wellbore (node’s) k nearest neighbors, in the topological approach, the algorithm draws an edge if and only if a wellbore (node) is within the induced fracture radius of influence of another wellbore (node). That is – if the induced fracture radius of influence of wellbore nodes overlap, then an edge is drawn. It is apparent from the results that the knn algorithm is, by definition, sensitive to the number of neighbors, with the edge connections and overall graph connectivity varying greatly based on this value. It is further apparent that even when only one nearest neighbor is selected, the geometry of the edge connections is considerably different than the geometry of the edges in the topologic approach.

4.3.2 Armstrong PA: Testing the Algorithms Against a Known Leakage Scenario In March 2008, a pressurization of surface casing in a newly drilled oil and gas well in Armstrong County, PA resulted in migration of fluids through two other producing wells [29], mediated by connectivity in the subsurface

Depth

Spatial Analysis of Induced Fracture Networks 89

0 0.2 0.4 0.6 0.8 1 1.2

1.2 1

1.2 0.8

1 0.8

0.6

1 Long

0.5 0 0

(a)

0.6 0.2 0.4 Lat

0.8 1

0.6

0.4 Long

0.4

0.2

0.2

Lat

0 0

(b)

1.2 1

1 0.6

0.8 0.4

Long

1

1 0.6

0.8

0 0

0.2

0.6

0.4

0.6 0.4

0.2

(c)

0.4

0.2 Lat

(d)

0.2 0 0

1.2

1.2

1

1 0.8

0.8

1.2 1

0.6

0.4

0.6 0.4

0.2 0 0

1.2

0.6

0.8

0.4 (e)

0.8

1.2

0.8

0.2

0.2 (f)

0 0

0.2

0.4

0.6

0.8

1

Figure 4.5 Representative graph output: Graph representation of wellbore points (n =50). Blue dots represent wellbores (vertices), red lines represent edges, green ellipsoids represent radii of influence (a) topological approach in 3D; (b) topological approach in 2D form; (c) topological approach in 2D form (d) knn approach (k = 3) in 2D form (e) knn approach (k = 2); (f) knn approach (k = 1). Graph axes are represented non-dimensionally.

fracture network. Like the Bradford field, Armstrong County has an extensive regional history of oil and gas exploration. To test the performance of the graph-based spatial analysis on a real world leakage scenario, both the knn algorithm (k = 1) and the topological algorithm were applied to a subset of wellbores near the pressurized well that caused the leakage event.

90

Hydraulic Fracturing and Well Stimulation

4.3.2.1 Data Sources The leakage event (“Dayton Investigation”) was identified from a Pennsylvania Department of Environmental Protection (PADEP) report on oil and gas well stray gas cases [29], and the wellbore associated with the event was located in the PADEP Office of Oil and Gas Management Compliance Report [30]. A subset of 77 Armstrong County wells from the NETL/DOE Pennsylvania wellbore dataset were spatially selected using ArcGIS, and exported to a .csv file for import to the Matlab program.

4.3.2.2 Results Both the knn (k = 1) algorithm and the topological algorithm were applied to the subset of Armstrong County data. Results are shown in 2D form in Figure 4.6. Both graph-based analyses identified a cluster of wellbores as associated with the leakage event: However, the knn algorithm erroneously identified subsurface connectivity between several more wellbores than were reported as being associated with the stray gas leakage. Overall, the

Pennsylvania wellbores Status of wellbores Active Dayton incident well

N

0

500

1,000 Meters

1,500

2,000

Figure 4.6 Location of Dayton Incident Well.

Plugged &/or abandoned Other Armstrong country, PA

Spatial Analysis of Induced Fracture Networks 91 knn algorithm (A) results in a qualitatively well connected graph, suggesting that there is extensive subsurface connectivity in the region, even with a k of 1. The wellbore associated with the leakage (black arrow) is shown to be connected to at least 6 neighboring wells via overlapping induced fracture networks. In contrast, in the topological algorithm (B and C), the graph is far less connected; and the affected wellbore is connected to two other affected producing wellbores. The topological algorithm also identified three other clusters of connected wellbores and fracture networks in the region.

4.4 Discussion The spatial locations of wellbores, as well as the geologic characteristics of the reservoir, place physical controls on the formation of induced fractures. Many fracture and fracture flow models do not honor wellbore locations and related engineered fractures; or, they consider only the spatial or geologic attributes of the wellbores or reservoir in stochastically generating random fractures. The importance of considering both spatial and geologic attributes in identifying areas at greater risk for overlapping induced fracture networks is highlighted by comparing the topological to a simple distance-based nearest neighbor algorithm. Once identified, these regions can be targeted for finer scale modelling, and used as part of a cumulative modeling strategy to constrain uncertainty in the subsurface. The results of the model are also of value as a standalone piece of data for the development of science-based wellbore drilling, injection, and risk management plans. As shown in Figure 4.5, a nearest neighbor approach is by nature sensitive to the selected k number of neighbors. The subgraphs produced by the knn algorithm contrast considerably from the subgraphs produced by the topological approach. In particular, the knn approach suggests far greater connectivity between nodes than does the topologic approach, particularly when higher values of “k” are chosen. Conversely, when low values of “k” are chosen, several potential connections between nodes may be missed, since the algorithm chooses the nearest neighbor of a node, and not its range of influence on other nodes. Since the purpose of this model is to identify probable regions of subsurface connectivity – and wellbores and wellbore clusters at greater risk for unwanted fluid migration – the knn approach gives a substantially less accurate representation of the spatial extents that are likely to contain overlapping induced fracture networks. First – the k is user defined, and the

92

Hydraulic Fracturing and Well Stimulation

algorithm output is greatly dependent on the user selection. Second – the knn algorithm does not take into account the geologically-based induced fracture radius in drawing edges: This results in a connected graph, which is falsely suggestive of subsurface connectivity, and hence, higher risk for fluid leakage through the subsurface and through wellbores. In neglecting to consider geophysical attributes, knn relies solely on wellbore neighborhoods, and not actual induced fracture networks in defining connectivity. In contrast, the topological approach considers both geology and wellbore spatial locations in assessing the probability of induced fracture network overlaps. Edges are drawn if and only if a wellbore falls within the radius of influence of a nearby wellbore. Hence, the subgraphs are representative of both geologic and wellbore networks, and give a more realistic (and in a sense, more conservative) representation of probable connectivity. Furthermore, unlike knn, the topological approach is not susceptible to user

(a)

(b)

(c)

Figure 4.7 (a) Results of knn algorithm on Dayton PA area wellbore (k = 1) (n = 77); (b) Results of topological algorithm on Dayton PA wellbores (n = 77); (c) Results of topological algorithm on Dayton PA wellbores, zoomed in on wellbores associated with leakage. Black arrows point to overpressurized wellbore. All results are shown in 2D form for clearer visual representation of results.

Spatial Analysis of Induced Fracture Networks 93 bias in the selection of neighbors: the algorithm itself automatically identifies the induced fracture radius and number of associated connections (edges), based on literature supplied values for lithologic fracture reach. When tested against a known leakage scenario, the topological algorithm identifies two producing wellbores associated with the leakage event (Figure 4.7b and 4.7c). Although the knn algorithm also identifies affected wellbores, it also falsely identifies several other nearby wellbores as being connected (Figure 4.7a). In addition, the overall knn graph is qualitatively connected, resulting in a likely erroneous (overly connected) representation of the subsurface induced fracture network connectivity, even with a k value of 1. In contrast, the topological algorithm identifies 4 clusters of connected wellbores (Figure 4.7b), including the wellbores associated with the leakage event. The overall topological graph is qualitatively less connected than the knn graph, and appears to provide a more realistic representation of the connectivity of wellbores and induced fracture networks. The example application serves to highlight the importance of the geologic factors in assessing regions with probable natural and engineered flow pathways.

4.4.1 Uses for Industry and Regulators There is an ongoing need in both industry and regulatory domains for science-based tools from which to develop risk management programs. This is particularly true in complex systems such as subsurface geologic systems, especially those impacted by hydraulic fracturing. The potential for subsurface leakage through induced fracture networks and existing wellbores is a concern to regulators, operators, and public stakeholders. The ability to provide better predictive tools for spatial regions or wellbores at risk for such events meets a critical need for the development of sound risk management strategies. The method presented in this paper can provide valuable information to stakeholders, and helps to reduce uncertainty inherent to these complex systems.

4.5 Conclusions Overlapping induced fracture networks between collocated wellbores may increase communication in the subsurface, and create the potential for unwanted fluid flow. The generation of induced fractures is greatly dependent upon the structural and lithological characteristics of local geology, which is often difficult to accurately characterize in the absence of costly

94

Hydraulic Fracturing and Well Stimulation

geophysical surveys. A robust, adaptable method for analyzing the spatial regions and wellbores at higher risk for subsurface induced fracture connectivity has been developed and presented. The result produced by the method is based on geologic data, and provides a sound basis for reduction of uncertainty inherent in subsurface systems. It is shown that the topological graph theory algorithm is a potentially powerful tool for rapid characterization of subsurface geospatial data. The topologic graph algorithm has several advantages over the knn algorithm: it is not susceptible to user error in the selection of a “k”. And, in accounting for geologic factors, it provides a physically based, and hence more realistic, assessment of probable subsurface connectivity. The algorithm has been successfully demonstrated using a real world leakage scenario. Because subsurface modeling efforts tend to occupy the realm of “big data,” the method increases modeling efficiency in two ways: First, the graph structures employed by the method allow for rapid computations involving big data sets; and second, the method can be used to identify spatial extents at greater risk for induced fracture network communication, and hence targeted fracture and fracture flow modeling. The method output can be transformed back into a geographic coordinate system, and/or integrated into existing fracture or fracture flow modeling software, as part of a cumulative modeling strategy for risk mitigation. The information provided by this approach can be used by regulators and industry in developing sound risk management plans related to hydraulic fracturing operations.

Acknowledgements This work was completed as part of National Energy Technology Laboratory (NETL) research for the Department of Energy’s Complementary Research Program under Section 999 of the Energy Policy Act of 2005. This research was supported in part by an appointment to the National Energy Technology Laboratory Research Participation Program, sponsored by the U.S. Department of Energy and administered by the Oak Ridge Institute for Science and Education. The authors wish to acknowledge the technical feedback from Kelly Rose, Russell Schwartz, Circe Verba, and Grant Bromhal. Deborah also thanks her family for their support, with special recognition to Miriam and Michael Miller.

References 1. D. Glosser, K. Rose, and J. Bauer, Spatio-temporal analysis to constrain uncertainty in wellbore datasets: an adaptable analytical approach

Spatial Analysis of Induced Fracture Networks 95

2.

3.

4.

5. 6.

7.

8.

9.

10.

11.

12.

13.

in support of science-based decision making. Journal of Sustainable Energy Engineering, in press (2016). State Oil and Gas Groundwater Investigations and Their Role in Advancing Regulatory Reforms a Two-State Review: Ohio and Texas, Groundwater Protection Council (2011). K.M Keranen, H.M. Savage, G.A. Abers, and E.S. Cochran, Potentially induced earthquakes in Oklahoma, USA: Links between wastewater injection and the 2011 Mw 5.7 earthquake sequence. Geology 41(6), 699–702 (2013). D. Soeder, S. Sharma, N. Pekney, L. Hopkinson R. Dilmore, B. Kutchko, et al., An approach for assessing engineering risk from shale gas wells in the United States. In International Journal of Coal Geology, Special Issue: Unconventional Natural Gas and the Environment (2014) (in press). W. Ellsworth, Injection induced earthquakes. Science. 341(6142), 1225942 (2013). J. A. Montague and G. F. Pinder, Potential of hydraulically induced fractures to communicate with existing wellbores. Water Resources Research 51(10), 8303–8315 (2015). M. TIngay, J. Reinicker, and B. Muller, Borehole breakout and drilling-induced fracture analysis from image logs. World Stress Map Project (2008). M. D. Zoback, D. Moos, and L. Mastin, Well Bore Breakouts and in Situ Stress, J. Geophys. Res., USGS Staff Published Paper, 90, B7, pp. 5523–5530 (1985). R.M Dilmore, J.I. Sams, D.B. Glosser, and K.M. Carter, Spatial and temporal characteristics of historical oil and gas wells in Pennsylvania: implications for new shale gas resources. Environmental Science and Technology 49, 12015–12023 (2015). National Energy Technology Laboratory, FracGen and NNFlow software v. 14.9, https://edx.netl.doe.gov/dataset/fracgen-and-nffowversion-14-9 accessed 4/11/2016. P. Valko and M.J. Economides, Propagation of Hydraulically Induced Fractures – A Continuum Damage Mechanics Approach. International Journal of Rock Mechanics, Mineral Science, and Geochemistry 31(3) 221–229 (1994). C. Zeeb, E. Gomez-Rivez, P. Bons, S. Virgo, and P. Blum, Fracture network evaluation program (FraNEP): A software for analyzing 2D fracture trace-line maps. Computers and Geoscience 60, 11–22 (2013). J. Pach, Geometric Graph Theory, Mathematical Institute of Hungarian Academy of Sciences, http://dcg.epf.ch/page-117408-en.html accessed 4/11/2016.

96

Hydraulic Fracturing and Well Stimulation

14. L. D. Cohen and R. Kimmel, Global minimum for active contour models: A minimal path approach. International Journal of Computer Vision 24(1), 57–78 (1997). 15. J. Pereira-Leal, A. Enright, and C. Ouzounis. Detection of functional modules from protein interaction networks. PROTEINS: Structure, Function and Bioinformatics 54, 49–57 (2004). 16. S. Yook, Z. Oltvai, and A. Barabasi, Functional and topological characterization of protein interaction networks. Proteomics 4, 928–942 (2004). 17. M. Alizadeh, Z. Movahed, R. Junin, R. Mohsin, M. Alizadeh, and M. Alizadeh, Finding the Drilling Induced Fractures and Borehole Breakouts Using Image Logs. Journal of Advanced Research 10(1), 9–30 (2015). 18. West Virginia Geologic and Environmental Survey, Interactive Mapping http://ims.wvgs.wvnet.edu/index.html (2015). 19. N.-N.G. Intellingence Agency of EUA, World Geodetic System 1984, WGS-84. p. 3 (1984). 20. W. Dong, C. Moses, and K. Li, Efficient k-nearest neighbor graph construction for generic similarity measures. WWW 577–586 (2011). 21. V. Garcia, E. Debreuve, and M. Barlaud, Fast k nearest neighbor search using GPU. 2008 IEEE Comput. Soc. Conf. Comput. Vis. Pattern Recognit. Work. 2, 1–6 (2008). 22. Y. P. Mack and M. Rosenblatt, Multivariate k-nearest neighbor density estimates. J. Multivar. Anal. 9(1), 1–15 (1979). 23. R. Hammack, G. Veloski, and J. Sams, Rapid Methods for Locating Existing Well Penetrations in Unconventional Well Development Areas of Pennsylvania, Unconventional Resources Technology Conference, 20–22 July, San Antonio, SPE-178558-MS (2015). 24. PA DCNR. The Pennsylvania Internet Record Imaging System/Wells Information System (PA*IRIS/WIS): Gateway to Detailed Oil and Gas Well Information Data http://www.pairis.state.pa.us accessed during 2013. 25. Pennsylvania Spatial Data Access Center, Oil and Gas Wells in Pennsylvania, http://www.pasda.psu.edu/ accessed during 2013. 26. R. Arnold and W.J. Kemnitzer, Petroleum in the United States and possessions: a presentation and interpretation of the salient data of geology, technology, and economics of petroleum in each state and possession treated according to the conventional major field divisions; Harper & brothers (1931). 27. US Geological Survey. Minerals Yearbook Digital Collection http:// digital.library.wisc.edu/1711.dl/EcoNatRes.MineralsYearBk.

Spatial Analysis of Induced Fracture Networks 97 28. Pennsylvania Spatial Data Access Center, Geology of Pennsylvania, http://www.pasda.psu.edu/ accessed during 2015. 29. Department of Environmental Protection, Stray Natural Gas Migration Associated with Oil and Gas Wells http://www.dep. state.pa.us/dep/subject/advcoun/oil_gas/2009/Stray%20Gas%20 Migration%20Cases.pdf Accessed during 2016. 30. DEP Office of Oil and Gas Management Compliance Report, inspection ID 1692786 http://www.depreportingservices.state.pa.us/Report Server/Pages/ReportViewer. aspx?/Oil_Gas/OG_Compliance accessed during 2016. 31. M. Brudy and M.D. Zoback, Drilling induced tensile well fractures: implications for determination of in situ stress orientation and magnitude. International Journal of Rock Mechanics and Mining Sciences 36, 191–215 (1999). 32. J.R. de Druezy, P. Davy, and O. Bour, Hydraulic properties of two dimensional random fracture networks following a power law length distribution. Water Resources Research 37(8), 2065–2078 (2001). 33. N.E. Odling, Scaling and connectivity of joint systems in sandstones from western Norway. J. Struct. Geol. 19, 1257–1271 (1997). 34. O. Bour and P. Davy, Connectivity of random fault networks following a power law fault length distribution. Water Resour. Res. 33, 1567– 1583 (1997).

Part 3 OPTIMUM DESIGN PARAMETERS

Fred Aminzadeh (ed.) Hydraulic Fracturing and Well Stimulation, (99–124) © 2019 Scrivener Publishing LLC

5 Fracture Spacing Design for Multistage Hydraulic Fracturing Completions for Improved Productivity D. Maity*, J. Ciezobka and I. Salehi Gas Technology Institute, Des Plaines, Illinois, United States

Abstract Over the years, productivity studies conducted for horizontal multistage completions have shown significant stage-wise variability. Optimizing such completions could hold the key to unlocking true value from shale reservoirs and improving well economics. Traditional hydraulic fracturing programs use the same fracture design along laterals without any consideration for changes in the reservoir and wellbore conditions. Methods using mechanical rock properties require expensive petrophysical logging data, while those involving use of drill cuttings can be highly resource intensive and time consuming. In this paper, we introduce a novel approach, which utilizes routinely available data such as measurements made while drilling and petrophysical data as available within a fracture spacing design framework. We validate our approach through application on multiple wells by comparing results from our workflow with post completion production logging. Finally, we highlight the potential advantages and pitfalls in our approach and present a roadmap for future implementation in different plays. Keywords: Artificial intelligence, hydraulic fracturing, fracture spacing, fracture design, drilling data

5.1 Introduction While the advent of multistage hydraulic fracturing in long lateral completions has revolutionized shale oil and gas production, the process still *Corresponding author: [email protected] Fred Aminzadeh (ed.) Hydraulic Fracturing and Well Stimulation, (101–124) © 2019 Scrivener Publishing LLC

101

102

Hydraulic Fracturing and Well Stimulation

lacks the robust understanding of what happens downhole within the reservoir and what leads to the significant variability in productivity from completed stages. A significant amount of work has gone into improving our understanding of these completions, and some of these observations have provided enough information to try and improve fracture spacing design. Traditional approach to completion design involves use of transient rate-time analysis to identify key design parameters (permeability and fracture half lengths) and using them to predict well performance [1]. Stage spacing is a critical design parameter, which is impacted by considerations of reservoir permeability [2], stress shadowing effects [3, 4], SRV considerations [5], and economic considerations such as net present value [6]. Other more elaborate techniques at optimizing stage spacing include use of microseismic data [7], and fracture network modeling [8] to name a few. In practical applications, what is desirable is to take a holistic view of the completion process and utilize as much data and analysis as possible for design [9]. While holistic design techniques are in place, most methods do not account for variability in reservoir and completion effects along the long laterals. From production logs and distributed acoustic and temperature sensing data, we know that many clusters show insignificant to no production, creating zones with very low productivity. This clearly indicates that the “one size fits all” approach creates sub-optimal fracture design, and this has been abundantly recognized by the industry [10]. We believe that while design issues (such as fracturing efficiency) are important; formation quality is critical, as sections with lower quality should have modified fracture density to provide for adequate drainage. Many novel approaches have been suggested in the past few years, which involve a thorough investigation of the reservoir properties and the geomechanical aspects of completion in shale formations. One recent example highlighting an engineered fracture spacing design approach uses characterization of reservoir and completion quality, which are used to predict proper stage placement [11]. Microseismic monitoring and other geophysical tools can also allow for improvements in design based on observations [12]. Generating pseudo logs for lateral sections of wellbore based on observations from the vertical pilots is an established technique [13] for understanding laterals and improving associated completions. Methods looking at fracability alone and utilizing stochastic optimization techniques have been evaluated [14]. The need for running wireline petrophysical logs for at least the vertical pilots, and the need for core analysis and correlation are well understood. These are not routinely available, and therefore, we felt there was a need for a technique that can be applied on any well to provide a quick optimal design suggestion based on the historic

Multistage Hydraulic Fracturing Completions 103 field data available for the play in question and the mud log data from the well under consideration. For this study, we wanted to devise a technique that can systematically distribute fracture stages for more effective drainage of the reservoir without the use of expensive wireline or logging while drilling data. What is unique about our completion design approach is the use of mud log data for completion design in the absence of any wireline petrophysical or geomechanical data from that area. This is expected to work reasonably well, provided a predictive model for rock properties can be developed. In our approach, we use a hybrid AI (Artificial Intelligence) based modeling workflow to predict geomechanical properties where stage spacing design utilizes mud log gas shows, as well as predicted geomechanical rock properties, within a predefined design framework.

5.2 Method For most new field development programs in unconventional plays, vertical pilot wells are drilled, cored, and logged in order to gain a robust understanding of the formation before completions can be designed. These pilots provide valuable insights into the rock, including the mineralogy, in situ stress state, organic content, lithology, porosity, etc. to name a few. This wealth of data can be used to predict the behavior of the well laterals drilled from the pilot. Using data from such pilots, we propose to design predictive models for reservoir properties, which can be obtained from mud log data alone. This allows for wider applicability of the design methodology without compromising on upfront well completion costs. We propose a hybrid neural network (Neuro-Fuzzy) workflow, which uses mud log data and geomechanical predictive models to design a fracture density model, which can then be used to place fracture clusters. Mud logs typically provide estimates for observed gas shows, Gamma ray for geo-steering purposes, and rate of penetration data. We understand that Gamma ray logs can provide indications of shale layers, which have higher natural radioactivity. Gas shows could indicate possible productive or non-productive zones and also potential naturally fractured zones. However, the observed gas shows are influenced by rate of penetration which in turn can be impacted by a multitude of factors, not all of which are due to reservoir conditions. The gamma ray tool is also influenced by the erratic drilling speeds and varying wellbore conditions encountered during drilling, in general. In order to develop the suggested design framework, we have to answer some important questions posed at the outset. Most importantly we want to

104

Hydraulic Fracturing and Well Stimulation

see how these parameters relate to zonal productivity potential, whether the impact is verifiable, and which parameters are needed for reasonable design solutions. To find these answers, we use available data from multiple wells from the Marcellus shale play, and verify through observations the necessary framework for the proposed design approach.

5.2.1 Impact of Natural Fractures The initial step is to understand how some factors may play an important role within our fracture spacing design framework. Marcellus and other shale plays are known to have varying natural fracture distributions, and depending on the in-situ condition as well as the properties of the injected fluid/proppant, these could significantly enhance the productivity of the stimulated well. In the Marcellus play, prior data suggests presence of natural fracture swarms as a result of local stress perturbations occurring over geologic timelines [15]. These natural fracture swarms are known to contribute significantly to overall production by providing additional surface area for gas to move from matrix to the connected fractures and eventually to the producing well. Identification of naturally fractured zones is a key element in accurate understanding of well behavior but this is not easy to achieve due to the need for use of indirect measurement techniques or proxies to identify the zones where the reservoir is fractured a-priori. While there are many available techniques for fracture characterization in reservoirs, we use available microseismic data from one of the wells under study (henceforth Well #1) to characterize fractures. This is made possible due to the ways in which hydraulic fractures interact with naturally fractured rock and the impact such interaction has on the final fractured rock volume in terms of fracture network complexity, fracture network dimensions, and magnitude distribution of the microseisms [16]. In this context, we look at two different properties evaluated based on the distribution of induced microseismicity associated with the hydraulic fracturing process. The first is the b value distribution, which is obtained from the Gutenberg-Richter law, providing the relationship between the magnitude of the seismic event and the total number of earthquakes in any given region and time period of at least that magnitude [17]. Higher b value is indicative of a larger portion of small earthquakes compared to bigger ones. Since in the presence of natural fracture swarms, many re-activations are expected, b values tend to be higher when hydraulic fractures interact with such zones [18]. In this study, we look at the overall distribution of events and their b value estimates for every completed stage and try to interpret post completion production logs. We expect zones showing

Multistage Hydraulic Fracturing Completions 105

Normalized measure

higher b value to be indicative of presence of natural fractures and therefore, they should correlate strongly with gas shows and production log data run for some of these test wells. Similarly, higher length-to-width ratio (or the ratio of the two principal dimensions of the microseismic event cloud) is indicative of an elongated perturbed zone with lower degree of complexity in the created fracture network. On the other hand, a lower lengthto-width ratio suggests a more complex network, which could be due to substantive interaction of the propagating hydraulic fractures with natural fractures [19]. Figure 5.1 shows length-to-width aspect ratio as mapped 1.5 Microseismic derived “b–value” 1 0.5 0

Normalized measure

(a)

1000

2000

3000

4000

5000

6000

1.5 Microseismic derived W/L ratio 1 0.5 0

(b)

0

0

1000

2000

3000

4000

5000

6000

1.5

Norm. L/W ratio

1

R² : 0.81132

0.5

0

−0.5 −0.5 (c)

0

0.5 Norm. b-value

1

1.5

Figure 5.1 Stage-wise distribution of microseismic derived (a) ‘b value’ and (b) length-towidth ratio for a study well. Subplot (c) shows cross plot of the two showing a strong positive correlation between the two parameters.

106

Hydraulic Fracturing and Well Stimulation

with borehole microseismic data and how it correlates with evaluated b values for the same stages. Figure 5.2 shows another example where b value has been compared with production logs and other relevant data to highlight these observations and how they correlate with mud log gas shows. We observe a poor correlation between b values and observed fracture density from image logs, which is expected as image logs are subject to interpretation errors (and so is b-value analysis). However, image logs only provide a snapshot of fractures at the wellbore unlike b-value, which defines the spatio-temporal seismicity distribution. We observe reasonable correlation between sections showing very high flow contribution and sections indicating highly complex fractured zones from b value distribution. Finally, we observe a reasonable correlation between production log and highly fractured sections of the reservoir, as well as a reasonably strong correlation between production log and high gas composition from mud log gas show data. The observed correlation between mud log gas shows and production log data over certain depth intervals has been observed for multiple wells and provides one element governing our stage/cluster spacing optimization workflow. Even though the correlation is not perfect, in conjunction with gamma log readings and rate of penetration data, a strong correlation between the observed production and modeled geomechanical properties governing production in shale reservoirs should be possible as it may take care of some of the outlier observations. Norm. measure

1.5

Norm. measure

(a)

Norm. measure

(b)

(c)

1 0.5 0

0

1000

2000

3000

4000

5000

6000

0

1000

2000

3000

4000

5000

6000

0

1000

2000

3000

4000

5000

6000

0

1000

2000

3000

4000

5000

6000

1.5 1 0.5 0 1.5 1 0.5

% G as

20

(d)

10 0

Lateral length (ft.)

Figure 5.2 (a) Microseismic derived ‘b value’, (b) image log derived fracture density, (c) mud log gas shows (red: 1 and blue: 0), and (d) production log for a study well. The properties (except production log data) are normalized between 0 and 1. The black insert in subplot (d) indicates the section of the wellbore where production log could not be run.

Multistage Hydraulic Fracturing Completions 107 Based on the observations, our design workflow involves utilizing relevant routinely logged data from mud logs (gas shows, rate of penetration, and gamma ray) and models for rock properties such as Young’s Modulus and Poisson’s Ratio using data from Well #1. These in turn are used to predict rock brittleness, which we correlate with another brittleness measure from lithological distribution to validate the brittleness function before actual use. This is because lithology has an impact on rock properties and therefore, rock brittleness can be considered a function of Young’s Modulus and Poisson’s Ratio [20]. Broadly speaking, Increasing Young’s Modulus or decreasing Poisson’s Ratio is indicative of more brittle formations. This modeled brittleness is then used in conjunction with gas shows to identify the optimal hydraulic fracture/cluster density along the lateral. The basic framework governing our design is to provide for more cluster density in regions susceptible to lower productivity behavior in order to improve overall production from the completed lateral. Based on this fracture density model, clusters are populated along the length of the lateral by honoring the background modeled density values.

5.2.2 Workflow Apart from the pre-completion drilling data, the proposed workflow requires some rock properties derived from specialty logging such as dipole sonic or spectral azimuthal gamma for representative lithologic layers. This data is typically available for a given field, especially new development felds where a single or multiple pilot wells are drilled before a full field development drilling program is implemented. This is necessary to model for the same properties based on routine mud log data by deriving the necessary models. The workflow involves the following steps: 1. Training wells are nominated based on availability of relevant specialty logging data and the exact position of the well in relation to various shale sub-layers. The original highresolution mud log data is inverted by passing through a low pass filter, and the filtered data is compared with the original so as to make sure that the univariate statistics show a reasonably good match. Multiple inputs are generated using multiple filters. 2. The filtered mud log data are used as input to design a feed forward back-propagation neural network model. The model is trained to predict desired output rock properties (Young’s Modulus and Poisson’s ratio). The model design includes the

108

Hydraulic Fracturing and Well Stimulation usual training, validation and testing phase. Care is taken to prevent over-fitting of the data as well as having a reasonable network in terms of size. We use a network with single hidden layer and 10:1 ratio for number of hidden layer nodes when compared with number of sample points. Models specific to defined lithological layers are tagged and stored for application. 3. For application of the design, data from the candidate well is identified, and it is segmented based on predefined lithological layers. Corresponding models are applied after careful filtering of data using the filters obtained in step 1. 4. Based on the generated geomechanical models, a rock brittleness parameter is computed. The resulting parameter is combined within the predefined hydraulic fracture density design framework shared later. It works on the basis of partially weighting mud log gas shows and modeled properties to get the final density values. The density model is normalized and fractures/clusters are placed based on the behavior of this modeled parameter.

5.2.3 Model Fine-Tuning We need to highlight that there are multiple input properties being used in the modeling process where each comes with varying degrees of associated uncertainties. As an example, modeled rock properties have a high degree of uncertainty due to modeling errors, particularly at large offset from design wells. Similarly, mud log gas shows can sometimes show erroneous readings due to gas flow into the wellbore downstream of the drilling bit. Similarly, the framework relating modeled rock properties and gas shows with naturally fractured zones in the reservoir is loosely defined due to a lack of adequate corroborative data. Moreover, multiple input sets are generated from singular properties using variable filter parameters. All these add up to create a highly non-unique solution space and therefore, finding the right framework for combining these parameters to define a hydraulic fracture density model can be a challenge. In order to tide over these uncertainties, we use a fuzzy classification technique to identify the definition boundaries with adequate fuzziness so as to classify sections of the lateral in terms of cluster spacing design by taking into account the underlying uncertainty as well. At the same time, if production logs are available and the broad framework is well defined (such as highly brittle rock and high gas shows should lead to a lower modeled hydraulic fracture spacing density, etc.), we can try to generate the best

Multistage Hydraulic Fracturing Completions 109 possible model (and correspondingly, the best possible fuzzy classifier) to match the designed fracture density with the observed production behavior post completion. This is accomplished by using an evolutionary algorithm to minimize a predefined error function which tries to match the inverse of modeled fracture density with the observed cluster wise production. The fuzzy rules set which defines the modeling framework we used is as follows: • Rule #1: Low modeled brittleness and low gas shows imply very high density. • Rule #2: Medium modeled brittleness and low gas shows imply very high density. • Rule #3: High modeled brittleness and low gas shows imply high density. • Rule #4: Low modeled brittleness and medium gas shows imply high density. • Rule #5: Medium modeled brittleness and medium gas shows imply medium density. • Rule #6: High modeled brittleness and medium gas shows imply medium density. • Rule #7: Low modeled brittleness and high gas shows imply low density. • Rule #8: Medium modeled brittleness and high gas shows imply low density. • Rule #9: High modeled brittleness and high gas shows imply very low density. Here the ‘density’ values indicate final fracture density (or perforation cluster density) recommendation to be made by the designed fracture density model. We do note that these rules suggest relatively lower fracture cluster count for the so called sweet spots. Since the decision on how much to frac and where is a highly complex one with well economics playing a major role, the workflow is adaptable enough so that the rules can be flipped with the high density recommendations changed to low density recommendations and vice versa. This approach is useful in cases where specific well intervals have a predictable behavior and sensitivity to stimulation.

5.2.4 Need for Artificial Intelligence There are three computational elements using AI techniques used in this work-flow. We need to consider the need for using said methods in this study. We understand that though broad relationships between Gamma

110

Hydraulic Fracturing and Well Stimulation

Ray measurements and rock properties are expected due to influence of clay content on said properties, the relation may not always hold due to other influences. The same holds true for hydrocarbon indicator used for modeling (mud log gas shows). Due to this non-linearity and in-exactness in the relationship between geomechanical properties and gamma ray, Artificial Neural Nets are ideally suited since they can map highly nonlinear relations if properly modeled and calibrated and are very robust in handling noisy data [21]. Furthermore, a broad correlation between modeled and observed properties and the desired application (hydraulic fracture spacing design) can be easily defined [e.g. higher gas shows and higher modeled brittleness leads to lower cluster density, etc.]. However with the high uncertainty in the available inputs, a simplistic framework for combining said properties may not capture the relationships accurately. A classic solution to such a classification problem is to use a Fuzzy Inference System. They are easy to understand as they are governed by fuzzy rules, which are semantic in nature, even though the underlying evaluation is mathematical. They have the ability to optimally search for the best classifier set definitions to match the observed data. They are simple yet highly adaptable and can work with the imprecise and incomplete data that we have [22]. Finally, as stated already, the designed fracture density and observed production behavior mismatch is minimized by using an evolutionary search routine. The big advantage with using such an approach is that it is highly scalable and adaptable and can be used to solve for multi-dimensional, non-differential, non-continuous and non-parametric problems. They are intuitive and very easy to build and therefore provide an optimal search algorithm for the problem at hand [23].

5.3 Data We apply this approach to three wells from two separate well pads (separated by 10’s of miles from one another). The data from Well #1 for Pad #1 is used as the design data as the well had open-hole logging carried out for the vertical pilot as well as the horizontal section of the well. Two other wells were used as application wells for validating the models as each had a production log available for independent validation. These include Well #2 associated with the same well pad as Well #1 and Well #3 associated with Pad #2. Figure 5.3 shows the various lithological layers of relevance across Well #1. We can clearly observe that the well lateral intersects two layers (Zone C or the target zone and zone B which is the overburden lower

Multistage Hydraulic Fracturing Completions 111 4900

Marcellus Onondaga Top cherry valley Top zone A Top zone B Top zone C Top zone D Well #1

5000

TVD (ft)

5100 5200

5300

5400

5500

5600 4000

5000

6000

8000 7000 MD (ft.)

9000

10000

11000

12000

Figure 5.3 Well #1 cross-section showing well track in reference to the varying lithologic formations in the Marcellus shale play. The lateral is restricted to Zones C & D.

Marcellus layer) of interest, and we use the available data to model the geomechanical properties for these two layers. Since we have extensive wireline logging done for this long lateral (Well #1), we can use the data from the logs to estimate geomechanical and other properties, which in turn form the basis for our model design framework using artificial neural nets. Apart from the standard mud log gas shows available from drilling records, other wireline tools run for this well (both for the horizontal lateral and the vertical sections) include standard measurements such as density, porosity, and resistivity, as well as lithological tools to identify mineralogy and organic content. A thorough petrophysical analysis was carried out for the entire logged wellbore, and the geomechanical properties were ascertained using the lithology data. Gamma from both the actual wireline logging run and the mud log data was correlated to validate applicability for other “application” wells which lack similar wireline logs. Using the geomechanical properties and the mud log data available for well #1, the entire dataset was pruned such that two separate datasets were generated. The inputs were expanded using multiple filtering bandwidths to extract features at different frequency spectrums, which might hold physical meaning and therefore are valuable in the modeling process. The inputs were in turn used to develop two separate models for two separate reservoir (shale) units, namely Zone C and Zone B as discussed earlier and observed from Figure 5.3. However, based on the well trajectory along the

112

Hydraulic Fracturing and Well Stimulation

lateral, the two sections have varied data density in terms of available sampling points. From the mud logs, while Zone C has approximately 600 data points, zone B only has approximately 70 data points, which makes results from the model defined for this zone susceptible to more errors. Figure 5.4 shows sample rock property (Poisson’s Ratio) modeled for these two layers, and we can observe relatively higher errors for the Lower Marcellus Layer compared to the target layer due to said mismatch. For network training, care is necessary to prevent either over-fitting or non-representative dataset generation. Care is taken in the neural network design with the ratio of number of network nodes to the number of data samples kept at less than 0.10. Also, segments from each Zone are combined making sure that they are representative of varying behavioral aspects of the property being modeled (such as sudden rise or drop in value). The final network models are chosen based on the minimized error observed within the network validation process (Figure 5.5). With the individual geomechanical models as well as the composite brittleness model ready for use, the next step is to identify the best fuzzy set definitions corresponding to the rules defined earlier. Figure 5.6 shows the final identified rules set which corresponds with the best match between the predicted fracture density models and the observed production log behavior for Well #1. Once the model is ready for application, all of the model design parameters are stored for later use with application test scenarios. These include Original Modeled

(a)

5760

5780

5800

5820

5840

5860

5880

5900

5920

(b)

5760

5780

5800

5820

5840

5860

5880

5900

5920

Original Modeled

(c)

5800

6000

6200

6400

6600

6800

7000

7200

(d)

5800

6000

6200

6400 6600 MD (ft.)

6800

7000

7200

Figure 5.4 (a) Modeled Poisson’s Ratio compared with actual log derived values and (b) corresponding error mismatch for Zone B and (c) Modeled Poisson’s Ratio compared with actual log derived values and (d) corresponding error mismatch for Zone C.

Multistage Hydraulic Fracturing Completions 113 1.0

Data Fit Y= T

Modeled property

0.9 0.8 0.7 0.6 0.5 0.4

R = 0.9074 0.3 0.3

0.4

0.5 0.6 0.7 0.8 Target property

0.9

1.0

Figure 5.5 Validation results from a sample neural network training run showing minimized error in modeled property (Poisson’s Ratio) compared to target property (actual log derived data). Note: The target data was normalized before modeling run.

Low

Medium

High

(in) (a) 0

50 in Low

Medium

100

High

(in) (b)

0

50 in Very low

Low

Medium High

100 Very high

(in) (c)

0

50 in

100

Figure 5.6 Optimal fuzzy set definitions based on training well data for (a) modeled brittleness, (b) mud log gas shows, and (c) output fracture density guidance.

114

Hydraulic Fracturing and Well Stimulation

the number of layers in the network, number of nodes in the hidden, input and output layers, trained weights associated with each node in the network, activation function s associated with each node, biases within the network, etc. These saved models are in turn applied to any dataset from “application” wells to generate hydraulic fracture/cluster density maps. The final fracture density maps need to be res-called in order to make sure that sufficient “maximum” and “minimum” fracture spacing is maintained before final fracture or perforation cluster placement. These maxima and minima limits can be independently evaluated using other stage design approaches discussed in the introduction to this study.

5.4 Results Before the models can be applied to other wells, they need to be categorized based on lithological layers associated with each portion of the wellbore to be analyzed. This is done by using the same approach as used for the training well (Well #1) as discussed earlier. Once the wellbore has been characterized, these segmented subsections are evaluated for rock properties by using the corresponding rock property predictive artificial neural net derived models. We will share the design results obtained using the models for the two application wells (Well #2 and Well #3) and compare the observations with production log data. This will allow validation of the observed results using independent production log results, which is critical as initial production is the key for rapid return on investment in shale gas wells. We again note that while Well # 2 belongs to the same pad as the training well (Well #1), Well #3 is located 10’s of miles from the first pad and incorporates a different completion design. Before we look into the application wells (Well #2 & Well #3), we apply the derived models to data from the training well itself (Well #1) using the segmented modeling approach. Based on the wellbore location in reference to lithological units (Figure 5.3), separate models are applied to the dataset along the wellbore. Figure 5.7 shows the results for this particular test case and as expected, we get a good match between the suggested fracture density derived from the proposed workflow and the production log data for Well #1. We note that the original completion for Well #1 involved 18 stages with 4 perforation clusters per stage and an inter-stage separation of ~280 feet. The modified spacing design as suggested by the workflow is shown in Figure 5.7(b). As observed from these results, significant mismatch is observed close to 6900 ft. (measured depth), which could be due to the wellbore lying either very close to or at the interface of zone B & zone C

Multistage Hydraulic Fracturing Completions 115 Production log inverse of designed density Poor match Good match

%age match: 82.1302

(a)

6000

6500

7000

7500

8000 MD(ft.)

8500

9000

9500

10000

10500

6000

6500

7000

7500

8000 MD(ft.)

8500

9000

9500

10000

10500

TVD(ft.)

5400 5450 5550

(b)

(c)

6000

7000

Designed clusters Confidence - low Confidence - high Cluster density - rescaled Cluster density - original

# Clusters/Stage: 4 # stages: 18 Smoothening Window: 40 ° of randomness: 0.01 Confidence: 50 Rescaling: 0.75

Equalspaced separation: 70 feet

7500

8000 MD(ft.)

8500

9000

9500

10000

10500

Figure 5.7 Figure showing (a) the comparison between modeled fracture density and available production log, (b) TVD behavior along lateral with traditional equally spaced perforation locations, and (c) the actual fracture spacing recommendation based on the model results for Well #1 incorporating both zone B & zone C models using segmented modeling approach. Red arrow indicates section with significant mismatch between production and designed fracture density.

(Figure 5.3). This would make it difficult to interpret which model is the right model and applicable for corresponding sections of the wellbore. For Well #2, we ran three tests with the first test incorporating the model associated with zone C for the entire lateral (Case 2A), the second test incorporating the model associated with zone B for the entire lateral (Case 2B), and the third test incorporating segment wise modeling using both models based on location of the lateral in relation to the lithological units (Case 2C). Figure 5.8 shows the modeling and design results for Case 2A, Figure 5.9 shows the results for Case 2B, and Figure 5.10 shows the results for Case 2C. We note that the original completion profile for Well #2 involved 14 stages with 4 perforation clusters per stage with an inter-stage separation of ~300 feet. We can clearly see sections along the wellbore where the design recommendation suggests sparser clusters and other sections which suggest denser cluster spacing. This correlates well with the production log results with sections suggesting denser clusters showing lower productivity and vice versa. This is desirable considering the defined modeling framework discussed earlier. However, for Case 2A & Case 2B, the predicted fracture spacing design does not match well with the observed stage-wise productivity behavior at some locations (identified by red arrow). This can be attributed to the model applicability issue in certain sections of the wellbore, depending on

116

Hydraulic Fracturing and Well Stimulation Production log inverse of designed density Poor match Good match

%age match: 87.2792

6000

6500

7000

7500

8000 MD(ft.)

8500

9000

9500

10000

6000

6500

7000

7500

8000 MD(ft.)

8500

9000

9500

10000

(a) TVD(ft.)

5400 5450 5500 5550

(b)

(c)

6000

6500

7000

Designed clusters Confidence - low Confidence - high Cluster density - rescaled Cluster density - original

# Clusters/Stage: 4 # stages: 14 Smoothening Window: 50 ° of randomness: 0.01 Confidence: 50 Rescaling: 0.75

Equalspaced separation: 78 feet

7500

8000 MD(ft.)

8500

9000

9500

10000

Figure 5.8 Figure showing (a) the comparison between modeled fracture density and available production log, (b) TVD behavior along lateral with traditional equally spaced perforation locations, and (c) the actual fracture spacing recommendation based on the model results for Well #2 incorporating model from zone C. Green arrow indicates section with significant mismatch between production and designed fracture density, which corresponds with wellbore section falling outside zone C.

Production log inverse of designed density Poor match Good match

%age match: 87.2792

TVD(ft.)

(a)

(b)

6000

6500

7000

7500

8000 MD(ft.)

8500

9000

9500

10000

6000

6500

7000

7500

8000 MD(ft.)

8500

9000

9500

10000

5400 5450 5500 5550

Equalspaced separation: 78 feet

(c)

6000

6500

7000

Designed clusters Confidence - low Confidence - high Cluster density - rescaled Cluster density - original

# Clusters/Stage: 4 # stages: 14 Smoothening Window: 50 ° of randomness: 0.01 Confidence: 50 Rescaling: 0.75

7500

8000 MD(ft.)

8500

9000

9500

10000

Figure 5.9 Figure showing (a) the comparison between modeled fracture density and available production log, (b) TVD behavior along lateral with traditional equally spaced perforation locations, and (c) the actual fracture spacing recommendation based on the model results for Well #2 incorporating model from zone B. Green arrows indicate sections with significant mismatch between production and designed fracture density, some of which corresponds with wellbore section falling outside zone B.

Multistage Hydraulic Fracturing Completions 117 Production log inverse of designed density Poor match Good match

%age match: 84.1999

TVD(ft.)

(a)

(b)

5400 5450 5500 5550

6000

6500

7000

7500

8000 MD(ft.)

8500

9000

9500

10000

6000

6500

7000

7500

8000 MD(ft.)

8500

9000

9500

10000

Equalspaced separation: 78 feet

(c)

6000

6500

7000

Designed clusters Confidence - low Confidence - high Cluster density - rescaled Cluster density - original

# Clusters/Stage: 4 # stages: 14 Smoothening Window: 50 ° of randomness: 0.01 Confidence: 50 Rescaling: 0.75

7500

8000 MD(ft.)

8500

9000

9500

10000

Figure 5.10 Figure showing (a) the comparison between modeled fracture density and available production log, (b) TVD behavior along lateral with traditional equally spaced perforation locations, and (c) the actual fracture spacing recommendation based on the model results for Well #2 incorporating both zone B & zone C models using segmented modeling approach. We do not see any significant mismatch between the predicted fracture density behavior and the productivity of completed perforation clusters.

whether the well track is within the zone defining the applied geomechanical model or not. For Case 2C, we observe a much better match along the entire completed lateral, and this is due to segmented modeling approach where the correct model (based on the location of the wellbore in reference to the lithological units) is used (Figure 5.10). This validates the applicability of the proposed approach for wells within proximity of the well used in training our models. For Well #3, the completion design was significantly different with 27 stages and 4 perforation clusters per stage with an inter-stage separation of 200 feet. Once again we generate results incorporating the model associated with zone C for the entire lateral (Case 3A), the second test incorporating model associated with zone B for the entire lateral (Case 3B) and the third test incorporating segment wise modeling using both models based on location of lateral (Case 3C). Since Well #3 is at an offset of 10’s of miles from Well #1, we expect the results to be not as robust as was the case with Well #2. Figure 5.11 shows the modeling and design results for Case 3A, Figure 5.12 shows the results for Case 3B, and Figure 5.13 shows the results for Case 3C. For Case 3A & Case 3B, the predicted fracture spacing design does not match well with the observed stage-wise productivity behavior at sections

Hydraulic Fracturing and Well Stimulation

118

Production log inverse of designed density Poor match Good match

%age match: 84.4068

8500

(a)

9000

9500

10000

10500

11000 MD(ft.)

11500

12000

12500

13000

13500

9000

9500

10000

10500

11000 MD(ft.)

11500

12000

12500

13000

13500

TVD(ft.)

8200 8300 8400 8500

8500

(b)

Equalspaced separation: 51 feet

(c)

8500

9000

9500

10000

Designed clusters Confidence - low Confidence - high Cluster density - rescaled Cluster density - original

# Clusters/Stage: 4 # stages: 27 Smoothening Window: 50 ° of randomness: 0.01 Confidence: 50 Rescaling: 0.75

10500

11000

11500

12000

12500

13000

13500

MD(ft.)

Figure 5.11 Figure showing (a) the comparison between modeled fracture density and available production log, (b) TVD behavior along lateral with traditional equally spaced perforation locations, and (c) the actual fracture spacing recommendation based on the model results for Well #3 incorporating model from zone C. Green arrows indicate sections with significant mismatch between production and designed fracture density, some of which corresponds with wellbore section falling outside zone C.

Production log inverse of designed density Poor match Good match

%age match: 84.6568

8500

9000

9500

10000

10500

11000 11500 MD(ft.)

8500 8500

9000

9500

10000

10500

11000 11500 MD(ft.)

TVD(ft.)

(a)

(b)

Equalspaced separation: 51 feet

(c)

12000

12500

13000

13500

12500

13000

13500

8200 8300 8400

8500

9000

9500

10000

12000

Designed clusters Confidence - low Confidence - high Cluster density - rescaled Cluster density - original

# Clusters/Stage: 4 # stages: 27 Smoothening Window: 50 ° of randomness: 0.01 Confidence: 50 Rescaling: 0.75

10500

11000 11500 MD(ft.)

12000

12500

13000

13500

Figure 5.12 Figure showing (a) the comparison between modeled fracture density and available production log, (b) TVD behavior along lateral with traditional equally spaced perforation locations, and (c) the actual fracture spacing recommendation based on the model results for Well #3 incorporating the model from zone B. Green arrows indicate sections with significant mismatch between production and designed fracture density, some of which corresponds with wellbore section falling outside zone B.

Multistage Hydraulic Fracturing Completions 119 Production log inverse of designed density Poor match Good match

%age match: 86.7141

8500

TVD(ft.)

(a)

9000

9500

10000

10500

11000

12000

12500

12000

12500

13000

13500

8300 8400 8500

8500

9000

9500

10000

10500

11000

11500

13000

13500

MD(ft.)

(b) Equalspaced separation: 51 feet

(c)

11500 MD(ft.)

8200

8500

9000

9500

# stages: 27 Smoothening Window: 50 Rescaling: 0.75

10000

10500

Designed clusters Confidence- low Confidence- high Cluster density - rescaled Cluster density - original

# Clusters/Stage: 4 ° of randomness: 0.01 Confidence: 50

11000

11500

12000

12500

13000

13500

Figure 5.13 Figure showing (a) the comparison between modeled fracture density and available production log, (b) TVD behavior along lateral with traditional equally spaced perforation locations, and (c) the actual fracture spacing recommendation based on the model results for Well #2 incorporating both zone B & zone C models using segmented modeling approach. Green arrows indicate sections with significant mismatch between production and designed fracture density.

highlighted using red arrows. This can be attributed to the model applicability issue in certain sections of the wellbore, depending on whether the track is within the zone defining the applied geomechanical model or not, as observed with the earlier test case. Other issues include the robustness of the model at separation of 10’s of miles as well as issues with inadequate data for zone B model as highlighted earlier. For Case 3C (Figure 5.13) using segmented modeling approach, we observe a much better match along the completed lateral. However once again, there are small sections of the lateral where the fracture placement recommendation based on density model does not match well with the production log observations. Next we highlight the issue of model robustness due to data inadequacy. As observed from Figure 5.12, a certain section of the modeled fracture density along the wellbore shows significantly poor results suggested by the consistent high values from ~9300 feet to ~11700 feet (measured depth). This is because the model from zone B is poorly defined due to lack of adequate data as discussed earlier. While the laterals for both Well #1 and Well #2 fall mostly within the target zone C, significant sections of the wellbore corresponding to the identified depth interval for Well #3 fall within overburden zone B (as observed in Figure 5.13 from ~10300 feet measured depth to ~11000 feet measured depth). Therefore, these erroneous artifacts

120

Hydraulic Fracturing and Well Stimulation

are observed for both Case 3B and 3C, which makes use of geomechanical models from zone B.

5.4.1 Applicability Considerations Based on our results, we can say with some degree of confidence that this approach can be useful in designing completions (stage or cluster spacing) of wells within the same well pad provided major sections of the wellbore do not fall very close to or at the interface between geologically distinct layers with significant variability in geomechanical properties. Moreover, presence of local faulting or completion of nearby wells post drilling operations of the candidate well can also have a significant impact on results. Since the proposed method works with multiple models, based on a segmented modeling approach, it is critical that each model is robust as well as well-defined, and its reliability should be ascertained before application. For wells at significant offset from those wells used for model design, applicability can suffer depending on lithologic variability across spatial distance, as well as other formation properties. However, our results show that properly designed models and segmented modeling approach can still provide reasonably good fracture density maps and spacing recommendations.

(a)

(b)

(c)

Figure 5.14 (a) Well schematic showing localized sweet spot due to intersecting natural fracture swarm (green box) and (b) hydraulic fracture density framework suggested in this study with (c) potential alternate framework along the wellbore lateral.

Multistage Hydraulic Fracturing Completions 121 Beyond the questions regarding effectiveness of the models, the methodology used, and applicability close to and away from those wells used for model training and design, a more fundamental question is the efficacy of the design framework proposed in this study. Figure 5.14 highlights the broad framework in question as well as one possible alternative framework to highlight this issue. Since the question of how to proceed with the completion design framework is a complex one with well economics playing an integral part in any decision making process, a more thorough investigation and decision making based on particulars of the wells being completed using this approach is essential. As an example, the decision on which framework to choose could be decided by price factors (gas vs. oil/condensate rich play) as well as reservoir related considerations (Clay richness, natural fractures, etc. to name just two).

5.5 Concluding Remarks We have introduced a fracture spacing design approach which makes use of routinely collected mud log data apart from some reference wireline specialty logs to model for complex geomechanical properties of the rock surrounding the well-bore. These models along with observed gas shows are used to propose variable perforation cluster spacing along the wellbore laterals. We have demonstrated this approach to be useful at small (100’s of feet) as well as large (10’s of miles) geographical offsets from the wells used to train said models. The proposed methodology identifies local sweet spots that require less stimulation and areas where more stimulation is needed. By redistributing the hydraulic fracture density along the wellbore, we aim to balance stimulation costs and long-term production performance. While many methods have been proposed over the years which utilize data from such specialty wireline logs to predict well behavior or recommend completion design, the key discriminator with our proposed fracture spacing design methodology is the ability to apply the technique at geographically far off wells without having to update the geomechanical models. The key is to use a well-defined modeling framework and production log or other completion quality attributes (such as from fiber-optic data) to constrain the designed models so that they can mimic well behavior upon completion with reasonable accuracy. Based on the results we have observed from multiple wells, including those shared in this study, we hypothesize that this approach should

122

Hydraulic Fracturing and Well Stimulation

work in most situations provided proper care is taken before applying this approach. However, the proposed approach needs to be validated as it may not hold under many situations depending on economic considerations. For future work, we propose to carry out extensive modeling studies and generate guidelines for applicability under varying scenarios as suggested in this work. In the future, we plan on using a fuzzy or probabilistic classifier to decide which model to be used depending on the closeness of the well track to a particular lithologic boundary. This is due to significant uncertainty ranging from 10’s to 100’s of feet when it comes to layer boundaries and exact well location, which can make the decision making on models to be used for design very non-representative. We hope to test this approach on multiple wells in other shale plays (Permian Basin). We also expect to conduct similar design work for multiple wells which are geographically spread out and validate these observations.

Acknowledgement This work was supported by Research Partnership to Secure Energy for America (RPSEA) project number 11122-20. We also acknowledge WPX Energy for providing access to active hydraulically fractured wells of opportunity in Marcellus shale play and Schlumberger for acquiring the wireline and production logs used in this study.

References 1. J.W. Crafton and D. Anderson, Use of extremely high time-resolution production data to characterize hydraulic fracture properties, Presented at the SPE Annual Technical Conference and Exhibition, San Antonio, Texas, 24–27 September, SPE-103591-MS, http://dx.doi.org/10.2118/103591-MS (2006). 2. P. Ye, L. Chu, I. Harmawan, et al., Beyond linear flow analysis in an unconventional reservoir. Presented at the SPE Unconventional Resources Conference, The Woodlands, Texas, 10–12 April, SPE-164543-MS, http://dx. doi.org/10.2118/164543-MS (2013). 3. K. Wu and J.E. Olson, Investigation of the impact of fracture spacing and fluid properties for interfering simultaneously or sequentially generated hydraulic fractures. SPE Production & Operations 28(4), 427–436 (2013). 4. M. Jonathan and J.L. Miskimins, Optimization of hydraulic fracture spacing in unconventional shales. Presented at the SPE Hydraulic Fracturing Technology

Multistage Hydraulic Fracturing Completions 123

5. 6.

7.

8.

9.

10. 11. 12.

13. 14.

15.

16.

Conference, The Woodlands, Texas, 6–8 February, SPE-152595-MS, http:// dx.doi.org/10.2118/152595-MS (2012). C. Cipolla and J. Wallace, Stimulated reservoir volume: A misapplied concept? Hydraulic Fracturing Journal 1(2), 54–75 (2014). X. Ma, T. Plaksina, and E. Gildin, Optimization of placement of hydraulic fracture stages in horizontal wells drilled in shale gas reservoirs, Presented at the Unconventional Resources Technology Conference, Denver, Colorado, 12–14 August, SPE-168790-MS, http://dx.doi.org/10.1190/URTEC2013-151 (2013). A. Baig, T.I. Urbancic, K.C. Mace, et al. Assessing the spacing of stages in plug and Perf completions through seismic moment tensor inversion, Presented at the SPE Hydraulic Fracturing Technology Conference, The Woodlands, Texas, 6–8 February, SPE-152547-MS, http://dx.doi.org/10.2118/152547-MS (2012). L.W. Bazan, S.D. Larkin, M.G. Lattibeaudiere, et al., Improving production in the Eagle Ford Shale with fracture modeling, increased fracture conductivity, and optimized stage and cluster spacing along the horizontal wellbore, Presented at the Tight Gas Completions Conference, San Antonio, Texas, 2–3 November, SPE-138425-MS, http://dx.doi.org/10.2118/138425-MS (2010). R.L. Kennedy, R. Gupta, S.V. Kotov, et al., Optimized shale resource development: Proper placement of wells and hydraulic fracture stages, Presented at the Abu Dhabi International Petroleum Conference and Exhibition, Abu Dhabi, UAE, 11–14 November, SPE-162534-MS, http://dx.doi. org/10.2118/162534-MS (2012). J. Ciezobka, Marcellus shale gas project, RPSEA, August 2013, http://www. rpsea.org/files/3629/ (accessed 11 May 2015) (2013). B. Ajayi, I.I. Aso, I.J. Terry, et al. Simulation design for unconventional resources. Oilfield Review 25(2), 34–46 (2013). N.R. Warpinski, M.J. Mayerhofer, E.J. Davis, et al., Integrating fracture diagnostics for improved microseismic interpretation and stimulation modeling, Presented at the Unconventional Resources Technology Conference, Denver, Colorado, 25–27 August, SPE-2014-1917906-MS, http://dx.doi. org/10.15530/urtec-2014-1917906 (2014). R. Borstmayer, N. Stegent, A. Wagner, et al., Approach optimizes completion design. The American Oil & Gas Reporter 54 (8), 79–88 (2011). A. Jahandideh and B. Jafarpour, Optimization of hydraulic fracturing design under spatially variable shale fracability, Presented at the SPE Western North American and Rocky Mountain Joint Meeting, Denver, Colorado, 17–18 April, SPE-169521-MS, http://dx.doi.org/10.2118/169521-MS (2014). T. Engelder, G.G. Lash, and R.S. Uzcategui, Joint sets that enhance production from Middle and Upper Devonian gas shales of the Appalachian Basin. AAPG Bulletin 93(7), 857–889 (2009). B. Bahorich, J.E. Olson, and J. Holder, Examining the effect of cemented natural fractures on hydraulic fracture propagation in hydrostone block

124

17. 18.

19.

20.

21. 22. 23.

Hydraulic Fracturing and Well Stimulation experiments. Presented at the SPE Annual Technical Conference and Exhibition, San Antonio, Texas, 8–10 October, SPE-160197-MS, http://dx. doi.org/10.2118/160197-MS (2012). B. Gutenberg and C.F. Richter, Seismicity of the Earth and Associated Phenomena, second edition, Princeton University Press, Princeton, NJ (1954). N. Boroumand, Hydraulic fracture b-value from microseismic events in different regions. Presented at GeoConvention, Calgary, Canada, 12–16 May (2014). I.A. Salehi and J. Ciezobka, Controlled hydraulic fracturing of naturally fractured shales: A case study in the marcellus shale examining how to identify and exploit natural fractures. Presented at the SPE Unconventional Resources Conference, The Woodlands, Texas, 10–12 April, SPE-164524-MS, http://dx. doi.org/10.2118/164524-MS (2013). R. Rickman, M.J. Mullen, J.E. Petre, et al., A practical use of shale petrophysics for stimulation design optimization: All shale plays are not clones of the Barnett shale, Presented at the SPE Annual Technical Conference and Exhibition, Denver, Colorado, 21–24 September, SPE-115258-MS, http://dx. doi.org/10.2118/115258-MS (2008). S. Haykin, Neural Networks: A Comprehensive Foundation, second edition, Prentice Hall, Saddle River, NJ (1998). T.J. Ross, Fuzzy Logic with Engineering Applications, third edition, John Wiley & Sons, West Sussex, UK (2009). D.E. Goldberg, Genetic Algorithms in Search, Optimization and Machine Learning, first edition, Addison-Wesley Longman, Boston, MA (1989).

6 Clustering-Based Optimal Perforation Design Using Well Logs Andrei S. Popa1*, Steve Cassidy1 and Sinisha Jikich2 1

Chevron North America Exploration and Production, Bakersfield, California, USA 2 University of Pittsburgh, Pittsburgh, Pennsylvania, USA

Abstract In an effort to better understand the well performance in one of the Chevron’s assets in San Joaquin Valley, a study was conducted to evaluate the perforation strategies and capture best practices. Well completion through perforation is typically performed using bare essential technology such as wireline logs and perforation guns. For basic reservoir formations, simple rules-of-thumb are used for perforation spacing and interval lengths. These are rarely validated by other methods, such as production logging and micro-seismic monitoring. For more challenging lithology, a more appropriate approach would be to place perforation clusters in target formations with similar properties. The research presents an efficient use of fuzzy clustering technology for identification of the optimum perforation strategy in a challenging waterflood diatomite reservoir. The methodology was applied on all newly drilled wells in the reservoir (within the last two years), and we found that this new approach improved our understanding over previous practices, not only by designing optimum perforations, but also an increased production was observed. Cluster analysis is the task of grouping a set of objects in such a way that objects in the same group (called a cluster) are more similar to each other (in some sense or another) than to those in other clusters. There are two commonly used types of clustering methods: hard and fuzzy clustering. In hard clustering, data is divided into distinct clusters, where each data element belongs to exactly one cluster. In fuzzy clustering, data elements can belong to more than one cluster, and associated with each element is a set of membership levels. The fuzzy clustering algorithm, also known as Fuzzy C-Mean (FCM) algorithm, was applied to log data of wells from different areas of a reservoir. Based on the clustering results, the workflow then identified whether the perforation was performed *Corresponding author: [email protected] Fred Aminzadeh (ed.) Hydraulic Fracturing and Well Stimulation, (125–140) © 2019 Scrivener Publishing LLC

125

126

Hydraulic Fracturing and Well Stimulation

on “good” regions (sand) or on “bad” regions (shale bedding). This information allowed the evaluation of the perforation jobs executed and also allowed capturing best practices and design changes for future well completions. The case study presents a simple, yet efficient workflow to extract additional information from logs and improve completion strategies and perforation design. The methodology is flexible and can be applied to any well where complex lithology creates a challenge in defining the optimum perforation intervals. Keywords: Fuzzy C-Mean, well perforation design, log analysis, hydraulic fracturing

6.1 Introduction The Lost Hills oil field is one of several significant oil fields located in the San Joaquin basin of central California [1]. The field is located about 45 miles northwest of Bakersfield in Kern County, Figure 6.1. The producing structure is a northwest-southeast trending double-plunging anticline [2]. The anticline is about 12 miles long and about 1 mile wide with high dipping flanks reaching up to approximatively ~20 degrees. The main reservoir is relatively shallow, and the producing interval varies from 600 feet to a maximum of 1200 feet along the crest of the anticline. The most productive unit is the Belridge diatomite of the Miocene

Figure 6.1 Lost Hills field location.

Clustering-Based Optimal Perforation Design 127 Monterey Formation. The diatomite is a biogenic siliceous deposit formed by the shells of diatoms with varying amounts of detrital material, principally clay and sand [3]. These deposits form a series of interbedded sequences within the entire productive interval, whereby thinner zones are sandier diatomite with better capacity for flow, while others are more compact with very poor permeability. Since its discovery in 1910, the reservoir was developed and produced in primary recovery. Hydraulic fracturing was initiated in the mid-1980’s and the technology proved to be a great avenue for increasing recovery given the natural low permeability of the reservoir. Waterflooding was introduced in the early 1990’s, initially localized and as a mitigation process to overcome the compaction of the highly compressive diatomite rocks. The water injection reduces compaction due to reduction of net effective stress on the reservoir rock. Compaction appears to be driving the crossflow from lower permeability to higher permeability layers and can be significant when adjacent layers have strongly contrasting permeabilities. The waterflood process proved to be very successful and contributed to additional recovery. As a result, the water injection was extended to the entire field. The initial development consisted of 2 ½ acre staggered patterns; however, with time, these patterns were infilled to 1 ¼ acre. After the 2000’s a relatively small part of the field was further infilled to 5/8 acre patterns.

6.2 Objective and Motivation The rapid and aggressive field development started in the late 1980’s when advances in hydraulic fracturing technology led to significant production increase and improved economics. Early on, technology employed multistage (plug and perf) hydraulic fracturing, followed by multi-stage coiled tubing fracturing technique, and after the mid 2000’s, use of Cobra jet/max fracs. Over the years the number of stages varied, back and forth, from a minimum of 5–6 to as many as 19 stages per well. The underlying reason was the full coverage of the hundred plus feet formation. However, the most interesting observation was the significant variance of the production performance of the wells after stimulation. The range of expected outcome was observed to vary anywhere from as low as 10–20 bopd to as high as 400–500 bopd at peak. Thus, multiple efforts were entertained over the years to try to establish a correlation between fracturing treatment attributes (including well design and completion) and well productivity. One of the most comprehensive hydraulic stimulation studies was performed in 2008 when more than 440 wells, completed between 1998 and 2007,

128

Hydraulic Fracturing and Well Stimulation

were analyzed to drive an optimum fracturing-treatment design to maximize production. The effort consisted of an overwhelming data collection exercise including cultural data, wellbore schematics, perforations, treatment by stage, additives by stage, pumping schedule, geological info such as formation markers (tops and bottoms), well production, and pattern production. While the study revealed some observations, it did not provide a silver bullet for frac completion optimization. The development of the field continues today with an active drilling program consisting of up to 50 wells per year. With every single producing well undertaking a hydraulic fracturing treatment prior to starting production, the opportunity to optimize the well completion and fracturing treatment is still of high interest. The main objective of the study is to investigate the impact of perforation placement and design on fracture execution and ultimately on the performance response of newly drilled wells in different development areas across the subject field. Additional motivation for the authors was to investigate if using only basic log data and clustering technologies reveals any hidden patterns, learnings, best practices and opportunities for future well completion optimization. It is a helpful benchmark exercise to analyze and contrast different perforations strategies driven by different lithology across the field.

6.3 Technology The well perforation strategy for this field was strongly linked to the number of completion frac stages planned for the well, and vice-versa. Therefore, a large number of frac stages, which would provide better frac coverage across the productive formation, would require an increased number of perforation sets. However, it was shown that one likely non-uniformity in well production is that fluid that is injected with intention to generate fractures from all perforation clusters is diverted to only a portion of the clusters leaving many clusters unstimulated [4]. Industry experience shows that there is an opportunity for improvement of recovery efficiency by stimulating effectively all perforation clusters along the well. Several approaches to improve stimulation uniformity have been proposed. In the “engineered completion” approach [4], well log data was used so that perforation clusters can be placed in regions of similar stress or brittleness index. Cipolla details new algorithms and an integrated workflow that could improve fracture treatment staging in both vertical and horizontal wells [5]. In a newer approach, Pierce and Bunger, promoted uniform stimulation of shale gas reservoirs by strategically locating entry points so that

Clustering-Based Optimal Perforation Design 129 stress interaction between the hydraulic fractures promotes rather than suppresses their stimulation growth [6]. This approach was inspired by computational studies showing that the “stress shadowing”, or more generally stress interaction can lead to either suppression or enhancement of multiple fracturing growth. Thus, there is a need for geologists and completion engineers to work to establish the best perforation placement along the wellbore, such as what fracs would initiate and propagate, to offer the best fracturing cover. In the case of the studied field the main observation indicated that well performance across the entire field shows significant variability, yet this is not clearly explained by any past stimulation and/or reservoir studies. However, it appears that production outcome follows a characteristic distribution which can be divided into three categories of wells: poor, average, and good. This classification inspired the application of clustering technologies in an attempt to research the underlying nonlinear correlations between formation characteristics, perforation placement, and production response.

6.4 Clustering Analysis Cluster analysis was originated in anthropology by Driver and Kroeber in 1932 and introduced to psychology by Zubin in 1938 and Robert in 1939. The technology was famously used by Cattell in 1943 for trait theory classification in personality psychology. Cluster analysis or clustering is a generic name for a set of statistical or machine learning algorithms that aim to detect distinct groups in a sample of objects, these groups usually being called clusters. In contrast to discriminant analysis, the cluster analysis does not require a group structure known a priori. This specific characteristic makes cluster analysis attractive as an exploratory tool [7]. Clustering can be achieved by various algorithms that differ significantly in their notation of what constitutes a cluster and how to efficiently define them. Popular notions of clusters include: groups with small distances among the cluster members, dense areas of the data space, and intervals or particular statistical distributions. Clustering can, therefore, be formulated as a multi-objective optimization problem. The appropriate clustering algorithm and settings (for example, distance function, density threshold, or the number of expected clusters) depend on the individual data set and intended use of the results. Typical cluster models include [8]: connectivity models such as Hierarchical Clustering [9], centroid models such as K-mean Algorithm

130

Hydraulic Fracturing and Well Stimulation

[10], distribution models such as Expectation-Maximization (EM) algorithms [11] where multivariate normal distributions are assumed, density models such as Density-based spatial clustering of applications with noise (DBSCAN) [12], subspace models such as two-mode-clustering, and graph-based models [13]. A “clustering” is essentially a set of such clusters, usually containing all objects in the data set. Additionally, it may also specify the relationship of the clusters to each other. Clustering can be roughly distinguished as hard clustering, where each object belongs to a cluster or not, and, soft clustering where each object belongs to each cluster to a certain degree. In the next section, the Fuzzy C-Mean (FCM) algorithm [Bezdek, 1981; Nock, 2006] is reviewed.

6.4.1 C-Means (FCM) Algorithm In contrast to k-mean clustering, where each data point belongs to exactly one cluster, fuzzy clustering associates a membership level to each of the clusters for every data point. These membership levels indicate the strength of the association between that data element and a particular cluster. Fuzzy clustering is a process of assigning these membership levels, and then using them to assign data elements to one or more clusters. One of the most widely used fuzzy clustering algorithms is the Fuzzy C-Mean (FCM) algorithm. Similar to the k-mean algorithm described in the last section, the FCM algorithm attempts to partition a finite collection of n elements (x1, x2, ..., xn) into a collection of k fuzzy clusters with respect to some given criteria. Given a finite set of data, the FCM algorithm not only returns a list of k cluster means = {μ1, μ2, ..., μk}, but also a partition matrix W = wi,j [0,1], i = 1, ..., n, j = 1, ... k, where each element in matrix W represents the degree to which data point xi belongs to cluster j. The minimization objective function in the FCM algorithm is as follows: n

arg min s

k

i 1

j 1

wi , j x1

2 j

(6.1)

where:

1

wi , j k

x1

j

l 1

x1

j

2 m 1

(6.2)

Clustering-Based Optimal Perforation Design 131 Note that (6.1) is different from k-means algorithm’s objective function , with by the addition of the membership values wi,j and the fuzzifier m m ≥ 1. The fuzzifier m determines the level of cluster fuzziness. A large m results in smaller membership values wi,j, and hence, fuzzier clusters. In the extreme case that m = 1, the memberships wi,j converge to 0 or 1, which is the same as a crisp partitioning. In the current study, a value for m = 2 was assigned, as commonly used in the literature.

6.5 Methodology and Analysis As stated in the objective section, the goal of this study was to analyze the impact of perforations placement and design on the well production outcome. Therefore, the study focuses on researching the nonlinear “mapping” between the available logs, existing perforations along the 600+ feet of productive formation, and production outcome of the wells. The following section will describe the data considered in this study and the methodology.

6.5.1 Available Data The three main sources of data consisted of reservoir characterization (wireline logs), well completion (perforations depth and footage), and well production performance (peak rate, first 3 months, etc.). A total number of 132 wells were drilled in the last two years and had available data for analysis. A map of the field is presented in Figure 6.3. The field was divided in three areas, as shown in Figure 6.3. Ten wells were randomly selected from each area for analysis. The motivation behind the field division and well selection was to capture any potential trends associated with field stratigraphy and depositional environment. Additionally, the wells were located in areas with different waterflooding maturity. The map shows the active existing wells with cyan. In contrast, the newly drilled wells within the last two years are shown in black. In the reservoir characterization domain, six types of log data were collected for all the wells. These logs are listed below, together with a brief explanation. 1. Deep Resistivity Log (DRES). This represents the deepest reading resistivity log. It is a measure of the ability of the formation to carry an electric current. The deepest resistivity is often believed to read past any near wellbore contamination from drilling mud.

132

Hydraulic Fracturing and Well Stimulation

Figure 6.2 Reservoir structure.

2. Gamma Ray (GR). This is a measure of the natural radioactivity of the formation. The tools physically count particles emitted by the formation. These counts are then calibrated to an API scale. 3. Neutron Porosity (NPHI). This is a measure of the capacity of the formation to attenuate neutron radiation. It is most sensitive to the presence of hydrogen, either in the form of water or hydrocarbons. In “clean” formations it is easily correlated to the porosity of the formation. 4. Permeability Log (PERM). This is a transform measurement that uses other logs’ information and attempts to estimate the formation permeability. For accuracy and validation, it  needs to be calibrated to laboratory measurements of permeability taken from whole core plugs. 5. Porosity Log (PHIT). This represents the total porosity. This track is calculated from the other logs and is an estimate of the true pore space capable of holding fluids.

Clustering-Based Optimal Perforation Design 133 Area 1

Area 2

Area 3

Figure 6.3 Field map showing the division of the three zones.

6. Shale Volume (VSH). The volume of shale is the fraction of the rock volume that is shale. It is calculated from the other logs. It is a critical input into the calculation of total porosity and water (or oil) saturation. In addition to the reservoir characterization data, completion information such as the depth and total perforated footages was collected for each of the wells. The perforations data was merged with the log data and stored in the same repository table. A snapshot of the sample data containing the data for the study is shown in Figure 6.4. For every half foot, there is a binary indicator (column COMPL_OPEN) to represent whether perforations are present. The total perforation footage for each well is calculated using this track data. In terms of production performance, several production indicators were considered. The following indicators were collected for all the wells: cumulative lifetime production, peak daily oil production, first month average production, first 3-month average daily production rate, peak (max) 3-month daily production, etc. In this study, we present the results for the peak 3-month daily production as the output production indicator.

134

Hydraulic Fracturing and Well Stimulation

Figure 6.4 A snapshot of the sample data showing collected log tracks and perforation completion.

6.6 Applying the FCM Algorithm For the purpose of this paper, we present the results of the ten wells sampled from each of the three areas of the field (see Figure 6.1). The FCM algorithm was applied to cluster the log data for each well. Initially, all six log types were used in the exercise; however, it was found that the calculated tracks (PERM, PHIT, VSH) did not bring additional value to the assessment. The final cluster analysis was performed using only the three measured log tracks. While originally the data was grouped into three clusters, to achieve higher definition in the partition zones we used five fuzzy clusters in the study. Thus the linguistic representation of the clusters, which describe the quality of the reservoir, can be defined as: bad, poor, medium, fair, and good. With the clustering performed for each well, the process identified whether the perforations were shot in the “good” (sand) regions or on the “bad” (shale) regions. The following step process was employed to cluster the regions within the reservoir formation for each sampled well: 1. For every half foot, the DRES, GR, and NPHI log data was input into the FCM algorithm. Also, the measurements that were less than 1000 feet were disregarded as not being part of the reservoir. 2. Run the FCM algorithm on xi = [DRESi, GRi, NPHIi], where i represents each half-foot measurement, on all the available log data for the well. The number of clusters, k, was set to be 5 in all the experiments. 3. The mean of each of the five clusters was returned from the FCM algorithm. The clusters were ranked by the DRES

Clustering-Based Optimal Perforation Design 135 log, where the one with the least DRES log was labeled as the “bad” (shale) region and the one with the largest DRES was labeled as the “good” (sand) region. An example of the cluster means returned by the FCM algorithm is shown in Figure 6.5. 4. The results of the clustering allow for associating each well with a cluster by depth. The clustering results for two sampled wells are depicted in Figure 6.5. 5. The final step was to calculate the perforation footage in the cluster label “good.” The procedure presented is explained through the two plots presented in Figure 6.5 above. The continuous lines represent the DRES, GR, and NPHI log data every half foot, while the blue “dots” at the right of the graphs represent the footage perforation existing in the wells. Lastly, the

0

50

100

150

200

–100 –50 1000

1200

1200

1400

1400 Depth, ft

Depth, ft

–100 –50 1000

1600

50

100

150

200

1600

1800

1800

2000

2000

2200

2200

2400

0

2400 DRES GR NPHI Perforated

C1 C2 C3 C4 C5

DRES GR NPHI Perforated

C1 C2 C3 C4 C5

Figure 6.5 Clustering results by depth for two sample wells. In this figure, C1 represents the shale region, and C5 represents the sand region. The actual perforation regions were also labeled by the dark purple regions.

136

Hydraulic Fracturing and Well Stimulation

column to the far right represents the “cluster” track where every half foot is assigned to one of the five clusters. This study was carried out using only fuzzy clustering. Other methodologies, such as hard clustering of ants colony optimization, can be applied to cluster the data. The procedure presented is explained through the two plots presented in Figure 6.5 above. The continuous lines represent the DRES, GR, and NPHI log data every half foot, while the blue “dots” at the right of the graphs represent the footage perforation existing in the wells. Lastly, the column to the far right represents the “cluster” track where every half foot is assigned to one of the five clusters. It should be noted that current field practices were rather inconsistent and selection of the perforations was accomplished through a visual inspection of each log separately. Certain logs may show particular signatures in the “good regions”, however, this was not very intuitive. The clustering allowed not only for a better definition of the regions but also for corroboration with the perforation intervals and production response.

6.7 Results and Discussion Two plots were generated to investigate the correlation between perforation placement and production performance. Figure 6.6 shows the plot of 3-month peak production over the total perforation footages for all the sampled wells, while Figure 6.7 depicts the plot of 3-month peak production over the total perforation footages on the “good” regions (identified by the FCM algorithm) for all the sampled wells. In Figure 6.6, with two exceptions, most of the wells have about 40 feet total perforations, however their 3-month peak production ranged from about 200 BPD up to more than 1,600 BPD. No visible correlation can be observed from the plot. In contrast, in Figure 6.7, a clearly visible correlation can be observed between the 3-month peak production and the footage perforated in “good” regions. The correlation stays valid for all three areas of the field. Similar analysis was carried out and plots generated for the other production indicators: peak daily oil production, first month average production, peak (max) 3-month daily production. Generally the plots showed the same trends with the 3-month peak; however the interpretations were impaired of both data availability (missing) and variability. The field does not benefit from any simulation models due to its complex geologic earth model and comingled production. Thus, a history match could not be performed.

Clustering-Based Optimal Perforation Design 137 Peak 3-month production (bopd)

1800 1600 1400 1200 1000

Zone 1

800

Zone 2

600

Zone 3

400 200 0

0

50 100 Total footage perforated per well (ft)

150

Figure 6.6 3-month peak production vs. total perforation footages for all sampled wells.

Peak 3-month production (bopd)

1800 1600 1400 1200 1000

Zone 1

800

Zone 2

600

Zone 3

400 200 0

0

50 50 100 150 Footage perforated in "good" zones (C5) (ft)

Figure 6.7 3-month peak production vs. total perforation footages on good regions identified by the FCM algorithm.

To better understand the data driven findings, a complex 3D fracture simulation model (Fracpro) was run for the same well in two scenarios. The stress profile used was derived from the available sonic logs run in a test well in the field. Consequently, both reservoir and treatment design were kept the same and the only difference between the runs was the placement of the perforation cluster. The two model runs are depicted in Figure 6.8 and Figure 6.9 respectively. To the left of each display there is the stress profile, a shaded area characterizing the “good” cluster and the perforation placement. To the left one can see the fracture length and conductivity at the end of pumping treatment.

138

Hydraulic Fracturing and Well Stimulation Strat well 31.33 min 0.000 1.100 1.650

Cluster 5

2.200 2.750 3.300 3.850 4.400

Proppant coverage lb/ft2

0.550

4.950 5.500 6.285 m/sec 263

1244 TXSP psi

2226

100

200 Fracture Penetration (ft)

300

Figure 6.8 Fracture modeling. Proppant coverage at the end of pumping, perforations at 2680 ft.

Strat well 368.63 min 0.000 0.900 1.350

Cluster 5

1.800 2.250 2.700 3.150 3.600

Proppant coverage lb/ft2

0.450

4.050 4.500 6.285 m/sec 263

1244 TXSP psi

2226

100

200 Fracture penetration (ft)

300

Figure 6.9 Fracture modeling. Proppant coverage at the end of pumping, perforations at 2750 ft.

The fracture simulator provided an effective representation of the fracture conductivity in the case of the two scenarios. In the first scenario, Figure 6.8, the fracture modeling with the perforations at 2680 ft. show high fracture conductivity (cyan and red color on the right chart) over the “good” region at the end of treatment. In contrast, in the second scenario, when perforations are placed at 2750 ft. below the “good” region, the simulation shows poor or low fracture conductivity over the “good” formation at the end of treatment, Figure 6.9. That supports previous published findings, whereby the fluid injected with intention to generate fracture across all formation of interest actually initiates and creates high conductivity paths at the perforation cluster, potentially leaving the rest

Clustering-Based Optimal Perforation Design 139 of the productive formation likely unstimulated. The simulations support the fuzzy clustering analysis, which revealed that perforations placed in the “good” regions of the formations return the highest production output.

6.8 Conclusions The paper presents a simple, yet efficient methodology to define the “good” footage regions for potential perforation in complex and heterogeneous reservoirs using FCM clustering. The study summarizes a step-by-step procedure that can assist the geologist and completion engineers in designing the perforation plan. The methodology demonstrates that the perforation design can be improved if one uses all the log data available for the well (such as DRES, GR, and NPHI) as well as the additional information defining the behavior of the logs. Using Artificial Intelligence (AI) techniques such as fuzzy clustering analysis methods, the well log data is effective in identifying rock intervals with similar properties. The clustering techniques demonstrate an unmatched value by exploiting hidden information from multiple attributes and drawing knowledge to assist in practical engineering applications. The methodology is flexible and can be applied to any well where complex lithology creates a challenge in defining the optimum perforation intervals.

Acknowledgements The authors would like to thank Chevron North America Exploration and Production for allowing the publication of this work.

References 1. S.A. Graham and L.A. Williams, Tectonic, depositional, and diagenetic history of Monterey Formation (Miocene), central Sand Joaquin basic, California. AAPG Bulletin 69, 385–411 (1985). 2. J. Brink and T. Patzek, Lost hills field trial - incorporating new technology for reservoir management. SPE 77646-MS, SPE Annual Technical Conference and Exhibition, 29 September-2 October, San Antonio, Texas (2002). 3. M.S. Bruno and C. Bovberg, Reservoir compaction and subsurface subsidence above the Lost Hills Field, California, Proc. 33rd US Symp. On Rock Mechanics, AA Balkema, Rotterdam, pp. 263–272 (1992).

140

Hydraulic Fracturing and Well Stimulation

4. R. Slocombe, A. Acock, C. Chadwick, E. Wigger, A. Viswanathan, K. Fisher, and R. Reischman, Eagle Ford Completion Optimization Strategies Using Horizontal Logging Data” SPE-168693. Unconventional Resources Technology Conference, 12–14 August, Denver, CO (2013). 5. C.L. Cipolla, X. Weng N. Ond, T. Nadaraja, U. Ganguly, and R. Malpani, New algorithms and integrated workflow for tight gas and shale completions SPE-146872, SPE Annual Technical Conference and Exhibition, 30 October-2 November, Denver, CO (2011). 6. A. Peirce and A. Bunger, Interference fracturing: Nonuniform distributions of perforation clusters that promote simultaneous growth of multiple hydraulic fractures. SPE Journal 20 (April 2015). 7. F.W. Wilmink and H.T. Uytterschaut, Cluster analysis, history, theory and applications. Multivariate Statistical Methods in Physical Anthropology, pp. 135–175 (1984). 8. L. Kaufma and P. Rousseeuw, Finding Groups in Data: An Introduction to Cluster Analysis, John Wiley, New York, NY (1990). 9. T. Hastie, R. Tibshirani, and J. Friedman, The Elements of Statistical Learning, Springer, New York, NY (2009). 10. J.B. MacQueen, Some methods for classification and analysis of multivariate observations, Proceedings of 5th Berkeley Symposium on Mathematical Statistics and Probability (1967). 11. A.P. Dempster, N.M. Laird, and D.B. Rubin, Maximum likelihood from incomplete data via the EM algorithm. J. Roy. Stat. Soc., Series B, 39(1), 1–38 (1977). 12. M. Ester, H.P. Kriegel, J. Sander, and X. Xu, A density-based algorithm for discovering clusters in large spatial databases with noise, Proceedings of the Second International Conference on Knowledge Discovery and Data Mining (1996). 13. P. Drineas, A. Frieze, R. Kannan, S. Vempala, and V. Vinay, Clustering in large graphs and matrices. Machine.

7 Horizontal Well Spacing and Hydraulic Fracturing Design Optimization: A Case Study on Utica-Point Pleasant Shale Play Alireza Shahkarami* and Guochang Wang Saint Francis University, Petroleum and Natural Gas Engineering, Loretto, Pennsylvania

Abstract Recent drilling and hydraulic fracturing activities in the Utica-Point Pleasant shale play have recorded true vertical depth of over 13,500 feet. Drilling wells at this depth is very costly and challenging. With the current commodity pricing, drilling in such conditions becomes unaffordable. One immediate solution to the current low energy prices is optimizing well spacing to enhance hydrocarbon recovery and, thus, the commercial feasibility of the project. Horizontal well spacing constitutes a fundamental parameter for the success of a shale-drilling venture. Determining the optimum horizontal well spacing in shale reservoirs represents a challenging task because of the complexity of controlling factors. These factors can be categorized into three groups: geological, engineering, and economic. Geological modeling and reservoir simulation are the standard tools utilized in the industry to integrate these controlling factors. In this study, we employed these tools to perform sensitivity analysis of reservoir characteristics and future production optimization for a deep drilling case study in the Utica-Point Pleasant formations. We sought to find the optimum horizontal well spacing scenario as well as hydraulic fracturing design, in order to attain the highest net present value for 50 years of gas production. Our reservoir model represented a portion of Utica-Point Pleasant formations at the depth of 13,000 feet and the dry gas window. A commercial reservoir simulator was coupled with an optimization algorithm to reach the best solution with a minimum simulation cost. Although the outcome of our study is subjective to the chosen asset, the workflow provides a good example of horizontal well spacing and hydraulic fracturing design optimization. *Corresponding author: [email protected] Fred Aminzadeh (ed.) Hydraulic Fracturing and Well Stimulation, (141–157) © 2019 Scrivener Publishing LLC

141

142

Hydraulic Fracturing and Well Stimulation

Keywords: Inter-lateral well spacing, production optimization, utica shale, point pleasant formation

7.1 Introduction Inter-lateral spacing (well spacing) is one of the biggest decisions in a field development in unconventional shale plays. Not having optimized well spacing can either cost the operators a lot of money or leaves a lot of money on the table. The well spacing must be close enough to help create as much of the stimulated reservoir volume (SRV) as possible; however, it must be wide enough to minimize fracture interference [1] (well to well interference and “frac hits”) and over-capitalization in a field development [2]. Various tools such as rate transient analysis (RTA) and numerical simulations can be used to optimize well spacing [3, 4]. At the very beginning of a field development, little data is available; therefore, it is very important to run sensitivity analysis and uncertainty investigation to assess the effect of uncertainties involved in reservoir characteristics of the field performance. Once pressure and production history data are available, the model can be calibrated to insure all of the assumptions are practically feasible. Well spacing is impacted by a combination of three types of factors defining the geological characteristics, the engineering process, and the economic constraints. The geological and reservoir characteristics assure the quality of the reservoir and the hydrocarbon presence. The engineering design provides the deliverability of the well. Finally, the economic factors warrant that the engineering project leads to profit [5]. The most important parameters that must be heavily studied are matrix permeability, fracture half-length (impacted by completions design), reservoir properties, capital expenditure, operating costs, and hydrocarbon pricing. The hydraulic fracturing enhances the reservoir permeability and leads to more recovery. The fracture permeability can be upscaled for fluid flow simulation using Oda’s method [6, 7]. In a high commodity pricing environment, it would make more sense to place the wells closer apart; however, low commodity pricing will dictate farther well spacing. A rock that has high matrix permeability would be better suited for a larger well spacing, but a rock that has lower matrix permeability is better optimized in a tighter well spacing. When designing well spacing, it is crucial to take completions design into account. For example, if 1,000 feet of well spacing is designed for a field development, it is very important to design the completions job in a fashion that would yield 500 feet fracture half-length by pumping enough sand and water, along with optimizing stage spacing, cluster spacing, and perforations design.

Hydraulic Fracturing Design Optimization

143

Therefore, a big portion of achieving the desired well spacing is creating completion jobs that would guarantee the attainment of such goals. Deep dry Utica located in both Pennsylvania and Ohio has sparked the interest of a lot of the operating companies in the past two years due to its attracting geology, massive production volumes, and the advantageous time value of money created in the production of as much volume as possible in the first 3 to 5 years of the life of the well. The biggest challenges in developing deep dry Utica have been high capital expenditure, lack of production history and reservoir rock characteristics, and optimal management of pressure drawdown to avoid prop-pant crushing/embedment and geo-mechanical effects. All of these challenges will be resolved and understood with time as this has been the case for all of the other shale plays developed to date. In order to create long-term value for the shareholders, the well spacing must be selected based on a spacing paradigm that yields the highest NPV. Production volumes could be important to the strategic development of a company, but the single economic parameter that creates long-term value for the shareholder of a company is NPV [5]. In this study, the well spacing was optimized using NPV. Optimal economical well spacing is achieved by having a solid understanding of fracture half-length, conductivity, and matrix permeability, which are typically very hard to obtain in unconventional shale plays and reservoir modeling. Therefore, various sensitivities must be run to understand the impact of these variables. In this study, we utilized a numerical reservoir simulation in order to model a dry gas reservoir 13,000 feet deep into a formation that resembles the Utica-Point Pleasant play. Initially by drilling one well, we studied the recovery after 50 years of production over a constant drainage area. Next, we looked for increasing the recovery factor by adding multiple wells. Finally, we wanted to find the optimum number of wells that leads to the highest NPV over 50 years of production.

7.2 Methodology 7.2.1 The Base Reservoir Simulation Model Our reservoir model covers a 1,003 acres area. For the sake of comparison, we kept the drainage area constant for the different well spacing scenarios. The drainage area is divided into 35,000 grid blocks, and the grid block size is 50 feet by 50 feet. The top of the reservoir is 13,000 feet deep, and the pay zone thickness is 200 feet, allotting into two layers of 100 feet thickness. These layers resemble the structure of Utica-Point Pleasant play. The reservoir contains dry gas, methane.

144

Hydraulic Fracturing and Well Stimulation

We used a commercial reservoir simulation package—CMG-BUILDER and IMEX [8, 9] to build and run the reservoir model. Table 7.1 lists the reservoir reference data for this study. The reference data was selected after consulting with multiple operators in the region. Figure 7.1 depicts a 3D view of the reservoir. In addition, Figure 7.2 shows the relative permeability Table 7.1 Reservoir characteristics of the base reservoir model. Base model reservoir characteristics Rock Compressibility

5.8e-6 l/psi

@

Reservoir Temperature

175

°F

Gas Density at Standard Condition

0.58

Air = 1

Pressure

12,350

psi

Depth

13,000

ft

Gradient Pressure

0.95

psi/ft

Water-Gas Contact (DWGC)

14,000

ft

3000 psi

Reference Pressure and Depth

Grid top (ft) 2015-01-01

13,100 13,090 13,080 13,070 13,060 13,050 13,040 13,030 13,020 13,010 13,000

Figure 7.1 Reservoir top 3D model built in CMG-BUILDER. The drainage area covers 1,003 acres and the reservoir includes 200 feet pay zone divided into two layers.

Hydraulic Fracturing Design Optimization

145

1.00 Krw vs Sw

Kr - relative permeability

0.80

0.60

0.40

0.20

0.00 0.20

0.36

0.52

Sw

0.68

0.84

1.00

0.65

0.80

0.95

0.80 Krg vs Sl

Kr - relative permeability

0.64

0.48

0.32

0.16

0.00 0.20

0.35

0.50

Sl

Figure 7.2 Relative permeability curves for this study. These curves are modified from the correlations available in the CMG-BUILDER.

curves utilized in this study. These curves are modified from the correlations available in the CMG-BUILDER. The base reservoir model has eight percent porosity and one micro-Darcy permeability. The wells are drilled and perforated in the bottom layer of the Point Pleasant formation. The length of the horizontal lateral is 7,000 feet. In

146

Hydraulic Fracturing and Well Stimulation

order to simulate the hydraulic fractures, we refined the grid blocks in a planar direction perpendicular to the wellbore. Figure 7.3 demonstrates an example of planar hydraulic fracturing design. The production is performed under the constant bottom-hole pressure of 2,000 psi for a period of 50 years, starting in 2015. The volumetric estimation of initial gas in place is 316.69 BCF. Table 7.2 summarizes the economic inputs required for the economic analysis study.

Well-S Well-3 Well-4 Well-E Well-S

Figure 7.3 Horizontal wells and planar hydraulic fracturing design.

Table 7.2 Economic analysis inputs. Economic analysis inputs Drilling Cost

$ 5,000,000

per 7,000 lateral

Hydraulic fracturing

$ 300,000

per stage

Royalty

20%

of Production

Operating Cost

$ 1.00

per MCF

Natural Gas Price

$ 2.50

per MCF

Yearly Discount Rate

10%

Hydraulic Fracturing Design Optimization

147

7.3 Optimization Scenarios In order to address the main objective of this study, well spacing optimization, we created five drilling scenarios in a constant drainage area (1,003 acres). Table 7.3 summarizes the number of wells and the corresponding well spacing selections. In addition to the well spacing, we wanted to find the optimum number of completion stages, and hydraulic fracture characteristics such as permeability, width, and half length. Table 7.4 shows the selected ranges for these parameters. Figure 7.4 demonstrates the workflow that we followed in this study. This work-flow comprises five general steps. The first two steps include building the base model and defining the ranges for the key parameters. Next, we create our objective function considering the field net present value (NPV) as the global goal to maximize. The process continues with coupling the reservoir model with an optimization algorithm. The optimization algorithm selects the values of the adjustable parameters from the Table 7.3 Five scenarios of inter-lateral spacing and the corresponding well spacing over a constant drainage area (1,003 acres) in this study. Scenario

Well spacing (ft)

#Wells

1

500

8

2

750

6

3

1,000

5

4

1,250

4

5

1,500

3

Table 7.4 Range of completion parameters. Parameters Hydraulic Fracturing Spacing

150–450

ft

Number of Stages

46–15

Fracture Permeability

1–10

Darcy

Fracture Width

0.0001–0.005

ft

Fracture Half Length

150–400

ft

148

Hydraulic Fracturing and Well Stimulation

Define the key parameters and their ranges

Build the base geological model

Define the objective functon for well spacing in shale reservoirs

Coupling the reservoir model to an optimization algorithm

NPV and well spacing optimization

Figure 7.4 The workflow for the well spacing and completion parameters optimization study.

range that the user has provided earlier. Then the optimization algorithm calls the simulator, and, in turn, the simulator runs the models and calculates the NPV values based on the economic analysis parameters. The output of this workflow is a collection of reservoir models, which allows us to select the optimum parameters based on the highest field NPV value. In this study, the whole process of optimization was carried out using CMG-CMOST [8].

7.4 Results and Discussion 7.4.1 Base Reservoir Model – A Single Well Case Figure 7.5 shows a scenario of the reservoir model with the base reservoir characteristics (Table 7.1) and just one well drilled. The figure shows the profile of gas rate (blue curve) and cumulative gas production (orange curve) for 1, 5, 10, 20, and 50 years after production starts. On the right side of each graph, the corresponding pressure distribution is also pictured. As it is clear from Figure 7.5, after 50 years of production, a significant portion of the reservoir is still untouched and does not contribute to gas production. This is a good example of how reservoir management methods may leave hydrocarbon behind. In addition, the total recovery factor in the scenario depicted is around 25%. Therefore, the need for drilling more wells is expected.

7.4.2 Multi-Lateral Depletion – Finding the Optimum Number of Wells Varying the well spacing, we designed five drilling scenarios on a fixed drainage area (Table 7.3). The optimization results suggest that the

Hydraulic Fracturing Design Optimization 1000

60000 100

50000

Cumulative gas SC (MMSCF) Gas rate SC (MMSCF/day)

40000 30000

10

20000 10000

1 2035 2045 Time, date

2055

2065

50 years

2025

20 years

5 years

0 2015

1 year

50 yrs

20 yrs

Gas rate, MMSCF/day

Gas production over 50 years 1 yr 5 yrs 10 yrs

10 years

Cumulative gas, MMSCF

80000 70000

149

Figure 7.5 The cumulative gas production, gas flow rate, and pressure distribution for different time steps (1, 5, 10, 20, and 50 years).

highest production belongs to 1,000 feet spacing between the laterals (Figure 7.6), which corresponds to the 5 wells. The second most prolific drilling scenario could be a 4 wells scenario with 1,250 feet spacing. In general, we judge the profitability of a project based on the NPV. Therefore, Figure 7.7 compares the optimum NPV for the drilling scenarios. While the 5 well scenario yields the highest NPV for this field, the 4 well scenario produces our second best NPV. The 4 well scenario could be an acceptable decision, particularly if there is limited capital budget for drilling and completion. In oil and gas development plans, short-term production could be vital in returning the initial investment. Figure 7.8 compares the recovery factor (RF) after 5, 30, and 50 years for all of the drilling scenarios. As it is shown in this graph, the 5 well scenario at 1,000 feet spacing has the highest recovery factor during all three time periods. In addition, considering just the 5 year recovery factor, a 6 well project built at 750 feet spacing could lead to a higher recovery factor than a 4 well scenario built at 1,250 feet spacing. However, we should also consider the limitations on capital budget. An important observation resulting from these scenarios is that almost all of them, except the 3 well (1,500 feet) scenario, yield more than a 50% recovery factor (almost two thirds of the ultimate production) after just 5 years of production. This is mainly because production of wells is limited to the stimulated reservoir volume.

150

Hydraulic Fracturing and Well Stimulation

Field cumulative gas production after 50 years Cumulative gas production, BCF

250 245 245.09

240 235

240.02

235.27

230 225

223.27

220 215

213.73

210 500

750

8

6

1000 Well spacing,ft

1250

5 Number of wells

4

1500

3

Figure 7.6 Field cumulative gas production after 50 years for five drilling scenarios.

NPV optimization for 50 years of production 140.0 127.4

130.0

119.1

NPV, $M

120.0 108.7

110.0

96.0

100.0 90.0 80.0 74.3 70.0 500

8

750

1000 Well spacing, ft

1250

1500

6

5 Number of wells

4

3

Figure 7.7 The results of net present value optimization study for five drilling scenarios.

Figure 7.9 compares the capital and operating costs of these projects with the NPV values. As it is expected, a higher number of wells will lead to a higher capital cost. However, completion projects typically require a significant share of capital cost. In addition, a higher production will increase the operational cost.

Hydraulic Fracturing Design Optimization Field recovery factor over different time periods RF after 50 years RF after 30 years

151

RF after 5 years

100.0% 90.0% 74.3%

Recovery factor (RF), %

80.0%

78.3% 70.3%

79.9%

77.0% 71.2% 65.5%

66.3%

70.0%

59.0%

54.3%

60.0%

81.6% 80.0%

75.5%

56.4%

50.0%

42.7%

40.0% 30.0% 20.0% 10.0% 0.0% 500

750

1000

1250

1500

4

3

Well spacing, ft 8

6

5 Number of wells

Figure 7.8 Field recovery factor (RF) over different time periods for all the drilling scenarios. Capital cost, $M

Operating cost, $M

210 180

(194.4) (162.3)

(174.7)

(172.7)

$M

150

(136.2)

127.4

119.1

108.7

120 90

Optimum NPV, $M

(88)

74.3

96.0 (67)

(66)

60

(53.6)

(40.2)

30 0 500

750

1000 Well spacing (ft)

1250

1500

8

6

5 Number of wells

4

3

Figure 7.9 Comparison of capital and operating costs with NPVs for different drilling scenarios.

7.4.3 Completion Parameters Table 7.5 summarizes the outputs of the optimization process for the completion parameters. Based on these results, a higher fracture conductivity constitutes the best solution. However, the fracture half-length is regulated by the well spacing, as expected. For instance, for the well spacing values

Well spacing (ft)

500

750

1000

1250

1500

Scenario

1

2

3

4

5

3

4

5

6

8

# Wells

Table 7.5 Optimized completion parameters.

250

250

250

350

350

Fracture spacing (ft)

28

28

28

20

20

# Stages

50

50

48.7

47.7

50

Fracture conductivity (md-ft)

400

400

400

350

250

Fracture half length (ft)

152 Hydraulic Fracturing and Well Stimulation

Hydraulic Fracturing Design Optimization

153

of 500 and 750 feet, fracture half-lengths are 250 and 350 feet, respectively. Obviously, higher values of fracture half-length will cause interference between the fractures (“frac hit”) and loss of injected fluids. The number of fracture stages depends on the space between the stages (fracture spacing). In addition, the cost of completion has a direct relationship with the number of stages. Therefore, an economic optimization analysis is required. We could suggest a higher number of stages for a lower number of wells per a constant drainage area and vice versa.

7.4.4 Second Economic Scenario, Reducing the Cost of Completion The optimization process is driven by the commodity costs. We created another scenario by reducing the cost of completion from $300,000 to $200,000 per stage. Figure 7.10 compares the NPV values for the two scenarios of completion costs for all drilling cases. As it is clear from this image, over the 50 years of production, we observe almost a constant shift (11–12% of NPV) between these scenarios. An interesting refection results from comparing the recovery factor values after 5 years of production for the two cost scenarios depicted in Figure 7.11. For well spacing cases of 750 and 1,000 feet, the reduction of completion costs leads to a recovery factor increase of 4.6% and 2.4%, respectively. These changes could be pivotal in decision-making, since in

NPV optimization for 50 years of production HF cost scenario 2 HF cost scenario 1 150.0

141.3

140.0

131.3

NPV, $M

130.0

122.5 127.4

120.0 110.0

108.0 199.1

108.7

100.0 90.3 90.0

96.0

80.0

74.3 70.0 500

750

1000 Well spacing, ft

1250

1500

8

6

5 Number of wells

4

3

Figure 7.10 Comparison of NPV between two costs of completion scenarios.

154

Hydraulic Fracturing and Well Stimulation Field RF comparison after 5 years between two cost scenarios RF after 5 years-HF cost scenario 1

RF after 5 years-HF cost scenario 2

100.0% Recovery factor (RF), %

90.0%

4.6% RF increase

80.0% 70.0% 60.0%

54.3% 54.3%

59.0%

63.6%

66.3% 66.2%

2.4% RF increase 56.4% 59.0%

50.0%

42.7% 44.7%

40.0% 30.0% 20.0% 10.0% 0.0% 500

750

1000 Well spacing, ft

1250

8

6

5 Number of wells

4

1500

3

Figure 7.11 Field recovery factor comparison after 5 years between two completion costs scenarios.

5 years of gas production shale reservoirs could bear more than 70% of ultimate production. The main reason behind the increase of recovery factor values goes back to having a higher number of completion stages due to the reduction of cost (Table 7.6). However, for the well spacing of 500 feet, the even increase of stages from 28 to 35 will not impact the production level after five years. This might be because of the tight well spacing that already covers all of the reservoir volume.

7.5 Conclusion In addition to the reservoir quality and engineering factors, the commodity costs and capital expenditure guide oil and gas development plans. In this project, we utilized geological modeling and reservoir simulation to study the profitability of a shale dry gas asset in the Utica-Point Pleasant formations. Due to the very tight formations, horizontal drilling and multistage hydraulic fracturing were planned to increase the reservoir deliverability. Our reservoir model simulates a portion of Utica-Point Pleasant formations at 13,000 feet deep and 200 feet thick. The drainage area covers 1,003 acres. Our simulation results showed that due to the tight shale formation, drilling just one well would leave a significant amount of natural gas behind. We wanted to find the optimum number of lateral wells over

Well spacing (ft)

500

750

1000

1250

1500

Scenario

1

2

3

4

5

3

4

5

6

8

# Wells

200

200

250

250

350

Fracture spacing (ft)

35

35

28

28

20

# Stages

49

49.5

48.2

36.8

47.8

Fracture conductivity (md-ft)

Table 7.6 Optimized completion parameters for the completion cost scenario of $200K per stage.

400

400

400

350

250

Fracture half length (ft)

Hydraulic Fracturing Design Optimization 155

156

Hydraulic Fracturing and Well Stimulation

the drainage area. By integrating the reservoir characteristics, economic data, and operational parameters the best drilling and completion scenarios were projected. Based on two different completion cost scenarios, $300,000 and $200,000 per stages, the optimum NPV for 50 years of production results from a 5 well drilling project. This number of wells includes 1,000 feet well spacing and 28 stages for each well. According to our optimization analysis, this project leads to a 77.4% recovery factor over 50 years. Drilling horizontal wells is a costly adventure and requires the availability of a significant capital budget. For this case study, a drilling scenario with just 4 well and 1,250 feet well spacing might be an attractive project. Although a 4 well scenario results in 6.5% less NPV compared to a 5 wells project, it will decrease the capital cost by 20%. In shale formations, the main part of production happens in the first couple of years. In our case study, almost 70% of ultimate production occurs in the first five years of production. This could be considerable since it can return the initial investment quickly. Our results demonstrate that a 5 well scenario gives the highest production after 5 years. However, changing the cost of completion can render the 4 and 6 well scenarios. Finally, although the outcome of our study is subjective to the chosen asset, the practice provides a good example of horizontal well spacing and hydraulic fracturing design optimization.

Acknowledgments The authors wish to extend their appreciation to Computer Modeling Group (CMG) for providing the software applications for reservoir simulation.

References 1. A. Awada, M. Santo, D. Lougheed, D. Xu, and C. Virues, Is that interference? A work flow for identifying and analyzing communication through hydraulic fractures in a multiwell pad. SPE J. (2016). DOI: 10.2118/178509-PA 2. R. Barree, S. Cox, J. Miskimins, J. Gilbert, and M. Conway, Economic optimization of horizontal well completions in unconventional reservoirs. PE Hydraulic Fracturing Technology Conference, The Woodlands, Texas, USA, Society of Petroleum Engineers (2014). DOI: 10.2118/168612-MS 3. A.S. Ziarani, C. Chen, A. Cui, D.J. Quirk, and D. Roney, Fracture and wellbore spacing optimization in multistage fractured horizontal wellbores: Learnings from our experience on Canadian unconventional resources. International

Hydraulic Fracturing Design Optimization

4. 5.

6. 7.

8. 9.

157

Petroleum Technology Conference. Kuala Lumpur, Malaysia: International Petroleum Technology Conference (2014). T.C. Frick, Importance of economics in production and reservoir engineering. J. Petrol. Technol. 10(9), 11–12 (1958). DOI: 10.2118/1026-G F. Lalehrokh and J. Bouma, Well spacing optimization in eagle ford. SPE/ CSUR Unconventional Resources Conference – Canada. Calgary, Canada, Society of Petroleum Engineers (2014). DOI: 10.2118/171640-MS M. Oda, Permeability tensor for discontinuous rock masses. Geotechnique. 35, 483 (1985). P. K. Ghahfarokhi, The structured gridding implications for upscaling model Discrete Fracture Networks (DFN) using corrected oda’s method. J. Petrol. Sci. Eng., 153, 70–80 (2017). Computer Modeling Group, Computer Modelling Group Manual (2015). C. L. Cipolla, E. P. Lolon, J. C. Erdle, and B. Rubin, Reservoir modeling in shale-gas reservoirs. Soc. Petrol. Eng. J. 13(4) (2010, August 1). DOI: 10.2118/125530-PA

Part 4 FRACTURE RESERVOIR CHARACTERIZATION Ahmed Ouenes

Introduction The role of discontinuities and the major impact they can have on the outcome of hydraulic fracturing has been recognized since the early days of field experiments carried by the Department of Energy and Gas Research Institute. Warpinski and Teufel [1] described many of these issues and attempted to model them with the best tools they had at their disposal at that time. Unfortunately, that initial effort was not continued, and the shale revolution was built on that initial knowledge and did not expand it significantly. With the oil crisis of 2014–2017, followed by financial pressures imposed by investors in 2018–2019 who were not informed about the risks of frac hits and well interreferences, the entire industry has to catch up on years of “natural fracture denial” where the hydraulic fracturing design of thousands of wells used software and technologies that does not include any representation of the natural fractures. New evidence from field experiments such as the one conducted at the Hydraulic Fracturing Test Site I industry consortium (HFTS-1) is again highlighting the importance of the natural fractures. The recent findings of HFTS-1 are well summarized in the September 2018 Journal of Petroleum Technology article titled “Real Fractured Rock is So Complex it’s Time for New Fracturing Models” [2]. 600ft of core taken in a hydraulically fractured Wolfcamp reservoir in the Permian Basin, USA showed a more complex reality than what is accounted for in most hydraulic fracturing design and analysis software. This includes the interaction between hydraulic and natural fractures, which is largely ignored or poorly accounted for in most software currently used to model hydraulic fractures and their resulting geometry. Rassenfos [2] emphasized in his summary of multiple recent publications describing HFTS 1 findings, that the “fracture height is overrated. While microseismic testing indicated that fractures grew up about a 1000ft, the height of the propped fractures- the fractures most Fred Aminzadeh (ed.) Hydraulic Fracturing and Well Stimulation, (159–180) © 2019 Scrivener Publishing LLC

160

Part 

likely to produce oil and gas was about 30ft.” Rassenfos summary article discusses the important role played by the natural fractures. The impact of natural fractures and discontinuities in general on hydraulic fracturing can be seen at different scales and manifest itself in different forms. In this section, four articles attempt to illustrate this impact while showing the variety of data that can be used to enhance our understanding and validate our new models. The four articles also illustrate the various numerical approaches which may be used to incorporate the effects of discontinuities. In Umholtz and Ouenes, the effects of faults and discontinuities is examined at a regional scale where local changes in stress created by a complex network of faults could inhibit or promote earthquakes resulting from stress perturbations created by deep water injection or hydraulic fracturing. Most of the past work in this area was limited to the conventional study by seismologists of fault stability which frequently was restricted to one or a limited number of these faults thus not capturing the true stress field created by a complex fault network. The authors used the Material Point Method (MPM) to incorporate a large number of faults over a regional scale to numerically compute the various stress properties that could be used to compute an Induced Seismicity Potential (ISP) that was validated using limited public data from Oklahoma. The use of numerical methods such as MPM applied to continuum mechanics augmented with discontinuities could provide the necessary means to capture the effects of the faults and fractures if adequate data is provided to model these critical discontinuities. Maity’s section shows the various data that can be used to capture the effects of natural fractures during hydraulic fracturing. In this article, some of challenges related to data collection and availability are highlighted. Maity studied two Marcellus wells that had not only microseismic data but also production and image logs which provided a quantification of the natural fractures at different stages. To quantify the performance of the hydraulic fracturing at each stage, Maity used a modified Nolte-Smith diagnostic tool to correlate different diagnostic parameters to production logs, microseismic and fracture density estimated from image logs. The microseismic data was quantified using the b-value. When plotting the various diagnostic plots against the production logs, b-value from microseismicity, and natural fracture from images logs, Maity found some weak correlations when using the production performance. However, stronger correlations were found between the b-value from microseismic with the extensional growth parameters compared to natural fracture interaction related parameters. The b-value showed no correlation with dilatational fracture swelling related parameters. Lastly, Maity observed no apparent correlation between the derived fracture density and the modeled

Part 

161

diagnostic parameters. This was expected since the image logs provide indication on the near wellbore region and cannot estimate the presence or absence of actual fractures in the formation. The author emphasized the multiple challenges associated with each type of data and the pitfalls that could result from not addressing data quality issues. Given all the challenges capturing the effects of natural fractures with the data that could possibly be available, Alzahabi et al. used a more engineering oriented approach to capture the effects of natural fractures and the resulting interaction with hydraulic fractures is homogenized into a Stimulated Reservoir Volume (SRV) with a certain average property. Given a certain SRV distribution estimated with simple geometrical properties of a fracture geometry, the goal is to find the best well positions using an optimization problem based on Integer Programming. Continuing into more computational approaches, Cai et al. use a semianalytical approach to predict the well performance in the case of refracturing. The authors use the concept of fracture-matrix cross-flow for a single fractured wellbore to derive their equations. The semi-analytical approach assumes some reservoir properties and focuses on the number of hydraulic fractures and their impact on the well productivity. Testing the approach against actual performances shows that the approach overestimates the well performance by 10%. The authors attribute this overestimation to the uncertainties in the reservoir parameters which will always remain an issue when using analytical and semi-analytical approaches. The sensitivity analysis conducted by the authors shows the increase in productivity as the number of hydraulic fractures increases but the benefits of such an increase levels out when the fracture conductivity reaches 2000 md-in. With these four articles, the reader is exposed to the variety of problems encountered in representing of the natural and hydraulic fractures and the consequences they have on the well performance. The important take away, is that multiple approaches exist and are in continuous development to create a toolbox that engineers can use to solve their many hydraulic fracturing challenges.

References 1. Warpinski, N.R. and Teufel, L.W., Influence of geologic discontinuities on hydraulic fracture propagation. J. Petroleum Technol., 39, 02, pp.209-220, 1987. 2. Rassenfos, S., Real Fractured Rock is so Complex it’s Time for New Fracturing Models, Journal of Petroleum Technology. 19 September 2018.

8 Geomechanical Modeling of Fault Systems Using the Material Point Method – Application to the Estimation of Induced Seismicity Potential to Bolster Hydraulic Fracturing Social License Nicholas M. Umholtz and Ahmed Ouenes* FracGeo, The Woodlands TX, USA

Abstract In an effort to promote awareness and understanding of the phenomena of induced seismicity, geomechanical modeling is applied to large publicly available datasets to demons trate the potential for bolstering social license. The Material Point Method (MPM) is used to simulate the interaction of fault systems with regional stresses. By combining mechanical results of the simulation to create induced seismicity potential (ISP) proxies, maps are generated to express areas of high and low inducement potential of seismic events. The results are compared to recent earthquake epicenter and injection well data. High coincidence of earthquake epicenters with regions of predicted high induced seismicity potential suggests the workflow presented could be deployed to quantify the risk of induced seismicity associated with the location of high-volume injection wells. The addition of a tool to assess the impact of location, not only injection volumes, is another critical step towards responsible regulation of injection wells, and mitigation of induced seismic events. Keywords: Induced seismicity, geomechanical modeling, material point method, Oklahoma earthquakes, Alberta earthquakes, social license, water disposal

*Corresponding author: [email protected] Fred Aminzadeh (ed.) Hydraulic Fracturing and Well Stimulation, (163–180) © 2019 Scrivener Publishing LLC

163

164

Hydraulic Fracturing and Well Stimulation

8.1 Introduction Recent media coverage has highlighted a major shortcoming in the industry of unconventional resource production, inducement of seismic events. Seismicity in certain areas of the midcontinent North America has increased drastically in the past few years. This has come to be understood as a result of pore pressure, and stress field perturbation due to water injection. Water disposal and hydraulic fracturing practices both introduce quantities of water to the subsurface which may alter the mechanical quasi-equilibrium established through geologic time. Such rapid perturbation of established stress fields can in some cases cause earthquakes. Because many of these operations are based in historically seismically quiescent regions, even magnitude 3–5 earthquakes may cause significant damage. The public’s perception and media’s coverage of this phenomena has produced a significant discussion topic, which impedes the rise of Shale 2.0, which will seek to achieve a more efficient development of shale resources. It is clear that operators do not understand the mechanical impact of such injections, as unpredicted seismic events have been induced at an increasing rate and of increasing magnitude. Regulators in Oklahoma have taken steps to reduce the risk of operators inducing seismic events, however the lack of means to quantify induced seismicity potential has hindered deployment of effective policy. The new tougher regulations imposed in Oklahoma are limited to reducing injection volumes and shutting down a limited number of water disposal wells around areas affected by recent earthquakes. In other words, the regulators are mostly reacting to the increasing earthquakes rather than proactively regulating the disposal process. With the advent of induced seismicity, institutionalized trust is compromised once again, and this time during a serious shift in the economies of unconventional production, underpinning not just the importance of the social license to operate, but the effect such public support can have when attempting to deploy the vast infrastructure necessary to realize unconventional petroleum resources at half of the value which they were previously produced. In order to address these concerns, a workflow is proposed, to combine the Material Point Method (MPM) and continuous fracture modeling (CFM) technologies with large geologically and geophysically constrained datasets. This workflow is applied to investigate the relationship between complex regional fault systems, and their collective impact on stress fields in a region. The mechanical outputs of this modeling workflow can be combined into proxies, and used to investigate areas of the model with stress fields more likely to be perturbed through high-volume injections

Geomechanical Modeling of Fault Systems 165 or hydraulic stimulation. Plotting these proxies as a continuous distribution on a map provides a practical tool that leads to very fast results that are simple to interpret, and which leverages large amounts of data and advanced modeling. This simple tool can be utilized by geoscientists, regulators, insurance companies, and the public, whose support for the petroleum industry is critical for the social license to operate.

8.2 The Social License to Operate (SLO) In Thompson & Boutilier [1], the concept of the social license to operate, and its direct relationship with the economic viability of an industry, is discussed. Traditionally, industries with large aerial, and temporal impacts on a community, such as mining, have been considered in such a light. Now, with the advent of unconventional resource production, the manufacturing approach of deploying hydraulic fracturing for unconventional petroleum resource production from shales (fracking) has drastically increased the footprint of the petroleum industry. This increased footprint elevates the potential for increased adverse effects as a result of industry presence and practices. Considered in this study is induced seismicity. In the pyramid of social trust, there are many boundaries which must be surmounted to approach an institutionalized state of trust on the part of the public (Figure 8.1). The public perception of the industry must evolve through legitimacy, to credibility, and finally trust, to provide the community support necessary to develop resources most efficiently.

Psychological identification Approval

Trust boundary Credibility boundary

Socio-political legitimacy

Acceptance Withheld/ withdrawn

Institutionalized trust

Legitimacy boundary

Interactional trust

Economic legitimacy

Figure 8.1 Diagrams showing the importance of credibility and trust in promoting social license. Critical to the highest level of public support, institutionalized trust, is sociopolitical legitimacy, and a transition from credible operation to trusted operations. Modified from Thomson and Boutilier [1].

166

Hydraulic Fracturing and Well Stimulation

In order to increase the public stakeholders in the industry, these boundaries: legitimacy, credibility, and trust, must be satisfied. In the current price environment, the economic legitimacy will be realized again in Shale 2.0 as new technologies drive production costs lower into a sustainable range. Interactional trust and the perception of socio-political legitimacy will only be established if new technologies can also serve to educate the public, and enhance regulators’ ability to ensure safe operations. If all of these conditions are satisfied, institutionalized trust can be established, maximizing public stakeholders’ engagement in unconventional resource production. The inability of regulators to stay ahead of such a significant industry impact as induced seismicity has cautioned landowners and the general public’s support of the unconventional petroleum industry. As volatile energy commodity prices, driven by geopolitical factors, continue to challenge the development of North American shale resources, it becomes exceedingly important to increase public stakeholders in the industry. Historically, there have been many setbacks to the public’s perception of the petroleum extraction industry. With the advent of hydraulic fracturing, a major concern which developed in the public’s eye was consideration of aquifer contamination. Sophisticated studies have been leveraged to both absolve most hydraulic fracturing from association with this contamination, and highlight shortcomings in completions, which have led to safer practices and more thorough inspections on the industry’s part to mitigate the hazard of aquifer contamination [2]. In a similar fashion, sophisticated multidisciplinary modeling tools are deployed in this study to continue a rigorous and scientific exploration of the potential drivers of induced seismicity.

8.3

Regional Faults in Oklahoma, USA and Alberta, Canada used as Input in Geomechanical Modeling

To address the issue of induced seismicity, collaboration between leading experts in many fields and disciplines is the only viable course of action. No single expert, whether they be hydrologically, seismically, or mechanically inclined, will provide a satisfactory way to address the issue. It is through the synthesis of these many disparate topics, that comprehensive models may be developed to help understand the ways in which humans may perturb evolving stress fields within the earth’s crust, and induce seismic events.

Geomechanical Modeling of Fault Systems 167 Oklahoma scientists and regulators have taken a stand to address this growing concern amongst their constituents. To this end, large, publiclyavailable datasets are being considered in numerous respects. Here, fault information (Figure 8.2) will be used as input in a geomechanical simulation that uses an advanced numerical model [3], the results of which will be compared to injection well and seismic epicenter data (Figure 8.3). Regulators in Alberta, Canada are also concerned, and have also hosted publicly available regional fault datasets, examined in Figure 8.4.

Oklahoma fault map

Figure 8.2 Fault map of Oklahoma [12]. These fault data are used as input to the geomechanical models simulated in this study.

Magnitude 3+ earthquakes 2006-present

Cumulative injection volume 2006–2013

Figure 8.3 Available earthquake and Underground Injection Control (UIC) well data from 2006-Present [13, 14]. Note the lack of earthquakes in multiple areas where large volumes were injected.

168

Hydraulic Fracturing and Well Stimulation

(c) (a) (b)

(d)

Figure 8.4 (a) Extent of the Montney/Duvernay shale play and considered area of interest, (b) Regional faults mapped by the Alberta Geologic Survey, (c) Faults in the study area between the cities of Fox Creek and Saint John where earthquakes of 4.2 and 4.5 magnitude, respectively attributed by Alberta Geologic Survey to hydraulic fracturing, were recorded in 2015, (d) major faults extracted for the geomechanical simulation and subjected to the regional stress.

Using these publicly available datasets, and new geomechanical modeling techniques, a workflow is proposed to develop fast industry tools to assist oil and gas companies, insurance companies, and regulators in addressing these public concerns. A geomechanical workflow which incorporates geological & geophysical data, with geomechanics [4, 5], is deployed on a regional scale to model the stress heterogeneities which exist as a results of the interaction between far-field regional stresses, and complex subsurface fault systems. Development and refinement of such a model on an oil or gas field level, where 3D seismic is available to accurately map all the existing faults, will provide a truly predictive tool to assess the risk of inducing a seismic event in any area where it is possible to leverage actual geologic and geophysical data.

8.4 Modeling Earthquake Potential using Numerical Material Models In attempting to develop a predictive model for seismicity, and especially induced seismicity resulting from water disposal or hydraulic fracturing,

Geomechanical Modeling of Fault Systems 169 the primary drivers of the phenomena must be characterized. An earthquake is the surface wave we experience standing on the earth as deformation equilibrates after stresses, built up through tectonic loading over time, release due to critical loading of a fault past a frictional restraint. It is therefore important to quantify the interaction between complex fault systems and regional stresses. What is a rock but a material, a material which to the ease of examination is populated with regular lattices, defined by varying minerals. The interfaces of the minerals, and in some rocks, the altered mineral interfaces, provide defects in the material where the potential to be separated is much higher than the atomic bonds within the mineral lattices. Through time, deforming the material will nucleate along the interfaces, and generate voids, or line deformations, which when given some amount of throw, in the very large scale are considered as faults. When considering the entire material, these defects will serve as a preferential nucleation zone for further deformation, especially when these fractures are in certain orientations with respect to the principle stress axes. A more complete discussion of the history of the study of failure in rocks can be found in Scholz [6]. It is therefore critical, for any model purporting to act in a predictive capacity, with regards to induced seismicity and seismic potential, to realistically capture the interaction of fractures (faults) with confining stresses, and furthermore, the interaction of multiple fractures with each other and such confining stresses. Successful prediction of the stress fields which arise from such interaction, will provide the ability to assess fault stability, and consequently seismicity potential within an evolving stress environment. Previous approaches to constraining the effects of earthquakes on subsequent earthquake potential have focused on the Coulomb failure criteria, when a Coulomb stress, Cf, exceeds a value

Cf = τβ + μ(σβ + ρ) where τβ is the shear stress on the fracture plane, σβ is the normal stress on the fracture plane, p is the pore pressure, and μ is the coefficient of friction along the fracture plane [7]. In these models, the Coulomb stress change, which is the difference between the Coulomb stress distribution before and after a major fault slip, highlights regions of stress perturbation, in many cases associated with subsequent seismicity or aftershocks. To better understand these complex stress fields and their relation to earthquakes on a field or regional scale, a new geomechanical workflow will be used and compared to previous results.

170

Hydraulic Fracturing and Well Stimulation

A major research topic that has preoccupied mechanical engineers for decades is modeling materials with discontinuities (cracks, fractures, faults). This is a critical research area since it is related to material failure which can cause major losses of life in areas such as aeronautics, and civil engineering. During the last decades major advances have been made in the study of crack propagation, crack intersections and material failure prevention. However, in most industries the number of fractures studied or modeled remain small compared to the problems found in the petroleum industry where rock failure is sometimes the goal. In hydraulic fracturing, the petroleum industry attempts to create a stimulated permeability by creating rock failure through hydraulic fracturing and leveraging the presence of natural fractures and the existing localized rock weakness where earth discontinuities have been formed throughout geologic time. When working at the scale of a well, this problem of studying the interaction between hydraulic fractures and a complex network of natural fractures poses a major engineering challenge. Modeling this large number of fractures and their propagation and interaction poses major computational and modeling challenges to classical numerical simulation methods such as finite elements. When modeling the geomechanical behavior of a large area that covers an entire oilfield or a producing trend the same problems arise since the number of faults present in the area could become overwhelming to most classical numerical simulation methods. To address this issue of handling a large number of discontinuities (natural fractures or faults) in geomechanical simulations, the Material Point Method (MPM) was introduced by Aimene and Nairn [8]. The introduction of the MPM method was accompanied with and continues to drive the ability to create new workflows that use quantitative data and information from geology, geophysics and geomechanics [4, 5]. The Material Point Method (MPM) is a meshless method developed as a potential tool for numerical modeling of dynamic solid mechanics problems [9]. It represents an alternate approach, with alternate characteristics, for solving problems traditionally studied by Dynamic Finite Element Methods. In MPM, a material body is discretized into a collection of points, called particles. It uses a background mesh as a computation space which allows the model to capture rapid and large deformation. Solid body boundary conditions are applied to the grid and/or on the particles. At each time step, the particle information is extrapolated to the background grid, to solve these equations. Once the equations are solved, the gridbased solution is used to update all particle properties such as position, velocity, acceleration, stress and strain, state variables, etc. This combination of Lagrangian (particles) and Eulerian (grid) methods has proven

Geomechanical Modeling of Fault Systems 171 useful for solving solid mechanics problems. Nairn [10] extended MPM to handle explicit cracks, resulting in the CRAMP method. By imposing discrete discontinuities in the model, CRAMP is able to simulate fractured materials using elastic fracture mechanics. The major inputs for MPM simulations applied to a study area are fourfold: 1) the distribution of rock geomechanical properties, such as Young’s Modulus, in the study area, 2) the distribution and properties of the discontinuities such as natural fractures or faults, 3) the distribution of the pore pressure, and 4) the boundary conditions representing the far field stresses acting on the study area. The MPM formulation is used to solve the classical dynamic continuum mechanics equations that include the momentum equation. The results of the MPM simulation is the usual stress and strain at the end of the dynamic simulation when the study area reached a quasi-equilibrium. More details on MPM and its use in the geomechanical modeling of multiple fractures and faults can be found in Aimene and Ouenes [3]. The MPM geomechanical computation was used to investigate a variety of geomechanical and engineering issues related to unconventional petroleum production at the well scale [3, 4]. Through several field validations, at the well and pad scales, this workflow has been demonstrated to successfully capture local stress variations, and predict complicated microseismic distribution in a number of stimulated wells. In this study, the workflow is now deployed on a regional scale inquiry to capture the interaction between large, complex, regional fault systems, and regional stresses. At the regional scale, the distribution of elastic properties and pore pressure used in the MPM simulation are considered constant. The discontinuities are represented by the interpreted regional faults. To test the validity of these assumptions at the regional scale, the results from MPM simulations are compared to previous studies conducted by seismologists studying the distribution of earthquakes near a fault. A simple case study using a boundary element method, from King and Cocco [7], is reproduced using MPM. The model presented by King and Cocco [7] plots a Coulomb stress change (Figure 8.5a), as a result of movement of the fault in the model. An approximation of Coulomb stress, captured by outputs of the MPM model, is compared to the distribution of stress in the original model, and is strikingly similar (Figure 8.5a,b). A proxy for Coulombs Stress (PCS) (Figure 8.5b) is defined as:

PCS = 0.5 * (σH − σh) * (sin(2β) − 0.4 * cos(2β)) − 0.5 * μ * (σH + σh) + μ * p

(8.1)

172

(a)

Hydraulic Fracturing and Well Stimulation

(b)

(c)

Figure 8.5 (a) Coulombs stress change as computed by King & Cocco [7], (b) An approximation of Coulomb stress computed in this work using MPM, (c) An Induced Seismicity Potential, ISP1, computed in this work using MPM.

where σH and σh are maximum and minimum principal stresses respectively, μ is the coefficient of friction, and β represents the angle between the considered fault and the compressive stress orientation. The goal of computing PCS and other possible Induced Seismicity Potential or (ISP) proxies is to demonstrate the possibilities for new quantifications of energy in the medium which may elucidate the regions most likely to express seismicity related to a fault in a specific stress field. Specifically, in this work a newly derived ISP1 is used,

ISP1 = |τ13| * |(σ11 − σ33) − (σH − σh)|

(8.2)

where τ13 is shear stress, σ11 is the maximum principle stress at the particle in the simulation, σ33 is the minimum principle stress at a particle, σH is the input maximum horizontal stress, and σh is the minimum horizontal stress input to the model. This quantity captures a particle’s subjection to shear stress and stress field perturbation due to the presence of a fracture. Plotting the proxy highlights areas near the fault that display an asymmetric distribution of aftershocks proximal to and oriented along the fault (approximately N-S), with quiescent regions surrounding the elevated occurrence of aftershocks, and with a secondary concentration of aftershocks distributed E-W further from the simulated fault (Figure 8.5c). Advances in processing power, and using MPM to simulate fracture mechanics, is desirable when investigating earthquakes as it allows for more realistic and comprehensive inputs to be considered by the model.

Geomechanical Modeling of Fault Systems 173 Advancing from models with only a single or handful of faults, a model is considered in this study which captures hundreds of regional faults, simulated in discrete sections, yielding over 2000 independent fracture planes to be simulated. An induced seismicity proxy, which is not dependent on calculations involving individual fracture plane orientation, greatly reduces the computational inefficiencies which have plagued previous inquiries.

8.5 A New Workflow for Estimating Induced Seismicity Potential and its Application to Oklahoma and Alberta For decades, the industry of petroleum exploitation has developed organizational structures that “siloed” professionals, which could have stymied multidisciplinary innovation. The ever-evolving economies of petroleum extraction continue to appreciate the value of data gathering, and especially the analysis and synthesis of those data into valuable information. To circumvent the shortcomings of siloing, and enhance the translation of value between geologists and geophysicists (G&G) and engineers, a 3G workflow that leverages the quantitative and simultaneous use of geology, geophysics and geomechanics was introduced to the industry [3, 5]. The use of a 3G workflow provides the fundamental framework within which to develop numerous subsidiary workflows to combine, enhance, and apply large G&G datasets with powerful mechanical modeling tools to address multiple issues related to unconventional resource production. Here, the workflow introduced by Umholtz & Ouenes [11] is applied to understand how large, regional-scale fault networks serve to perturb far-field stresses defined by tectonic loading. The resulting stress field attributes of such a model are then combined to generate an Induced Seismicity Proxy, which serves to identify regions of the model more or less likely to be perturbed by nearby or farther away large injection volumes. The resulting ISP predictions depend entirely on the preliminary input fault network, which is a work in progress [12]. Two areas in Oklahoma are of elevated interest to regulators (Figure 8.6), the area around the large population center of Oklahoma City, and an area slightly northwest, which has hosted an increasing number of seismic events since 2009. These will serve as the focus of the current study given the initial fault interpretation released by Oklahoma regulators [12]. Additionally, an area in Alberta, Canada (Figure 8.4), which does not yet demonstrate a significant number of seismic events, is modeled to demonstrate the workflow in a region where a few large induced

174

Hydraulic Fracturing and Well Stimulation

Simulation 2 Study area

Simulation 1

EFM EFM

Figure 8.6 Equivalent Fracture Models (EFM) derived from an area of interest in Oklahoma for input into geomechanical models.

seismicity events exceeding magnitude 4 have been attributed to hydraulic fracturing. With a powerful tool, able to capture the interactions between stress fields and complicated fracture systems, the next step in developing an induced seismicity proxy is selecting the appropriate outputs of the model to represent areas of preferential fault growth in the evolving stress field. As many authors modeling fault propagation have noted, Coulomb stress change appears to highlight areas of historical aftershock generation related to ruptures along large-scale faults. Building on these efforts, an alternative induced seismicity proxy is developed which combines both the shear strain and differential stress changes in the model as a result of tectonic loading on the regional fault system. This geomechanical proxy captures both the preferentially stored energy, represented by differential stress (σHmax − σhmin), and the shear strain on the particles (dv/dx and du/dy), capturing regions likely to accommodate slip along critically oriented planes.

ISP

dv dx

du dy

(

H max

h max

)

The results of the simulation using the preliminary fault interpretation of Holland [12] are summarized with the ISP proxy plotted in Figure 8.7. Regions of elevated ISP are warmer red colors, while areas of low ISP are cooler blue colors. Many of the regions of elevated ISP are found proximal

Geomechanical Modeling of Fault Systems 175

ISP High

Low

Figure 8.7 Result of the geomechanical proxy for ISP. Compared to historical earthquakes and underground injection control well data.

to fault systems, especially those critically oriented. Some regions of high ISP however, form conduits connecting less active areas of the fault network. Comparison to earthquakes data and injection well data can help substantiate these interpretations. Figure 8.7 compares the ISP proxy to earthquake data. We see that many of the earthquakes observed in Oklahoma fall into areas of high ISP, and many are located near critically oriented faults. Areas of low predicted ISP are often lacking in induced seismicity. Further comparison of the dataset, with consideration of injection well data, shows interesting features in low ISP zones. Namely, that some of the largest volumes injected, do not coincide with seismicity, and these instances are all in areas predicted as low ISP (Figure 8.8). For these reasons, the predicted ISP map (Figure 8.7) is a promising tool to help understand and regulate induced seismicity in Oklahoma by adding a missing critical component: location of the injection wells as compared to the geomechanical interaction between the regional stress and the regional fault network. Most regulatory efforts made to reduce induced seismicity potential are targeted at reducing injection volumes. These preliminary results indicate that the location of these injections may serve as a primary control over the expression of seismic events as the result of stress field perturbation through high volume injections. We also note that the preliminary fault interpretation of Holland [12] and its use in the resulting ISP shows some inconsistencies with recent major earthquakes (Figure 8.9). It is obvious from these recent earthquakes as well as previous earthquakes in that area that the initial fault interpreted by Holland [12] continues towards the North East and does not stop as shown in the preliminary fault interpretation. This observation shows the importance of accurately mapping all of the faults in high seismicity areas. This detailed mapping requires a statewide coordinated effort and can only succeed if all the seismic surveys and wells available in these areas are used. The proposed MPM geomechanical simulation and its resulting ISP

176

Hydraulic Fracturing and Well Stimulation

ISP High

Low

Low ISP regions Large injection volumes Limited earthquakes

Prague fault: high ISP Many earthquakes Limited injec

Figure 8.8 Zoom of Figure 8.7 that includes to the east the Prague fault, Oklahoma City and very large injection volumes in areas with no earthquakes. Notice that most of the earthquakes occur in high ISP areas and large volumes are injected north and west of the Prague fault in low ISP areas without causing any earthquake in their vicinity. Size of blue circles represent injected volumes. Size of red stars represent magnitude of earthquakes.

0

60 km

Figure 8.9 Zoom of Figure 8.8.3 to show seismicity NE of interpreted fault in Major County.

Geomechanical Modeling of Fault Systems 177 could be easily and quickly updated if a more complete fault interpretation becomes available in the future. The same process could be applied to Alberta where the faults shown in Figure 8.4d and their discrete representation shown in Figure 8.10a, could be used as input in the geomechanical workflow which will calculate the ISP shown in Figure 8.10b. The areas with red colors in Figure 8.10b could potentially have a higher probability of causing earthquakes as a result of water disposal or even large volume hydraulic fracturing. The implication of these results is that a predictive tool, which can be deployed over large areas, can be constrained in such a fashion that it is an understandable representation of the spatial variation of the potential to induce seismic events for regulators, E&P companies, insurance companies, and the public. Such a tool can be envisioned by considering the following maps for large areas in Oklahoma (Figure 8.11) and Alberta (Figure 8.12).

High Low

(a)

(b)

Figure 8.10 (a) Equivalent Fracture Model derived from the major fault map of Figure 8.2d (b) Resulting geomechanical output showing the complex distribution of the areas of high potential induced seismicity.

ISP High

Low

Figure 8.11 Derived ISP maps plotted along with injection volumes and earthquake data over a large area of the state of Oklahoma.

178

Hydraulic Fracturing and Well Stimulation

High Low

Figure 8.12 Derived ISP maps plotted in an area of Alberta where multiple large magnitude earthquakes attributed to hydraulic fracturing were recorded in Fox Creek and Fort St. John in 2015.

8.6

The Benefits of a Large Scale Predictive Model and Future Research

This study demonstrates the potential of MPM and its use in a 3G workflow to develop tools which may be valuable to regulators, insurance companies, as well as oil and gas companies in addressing induced seismicity risks. A workflow which incorporates geological and geophysical datasets, with powerful mechanical modeling techniques, was applied to develop a tool for understanding induced seismicity potential in a region. The benefits of the approach outlined in this study are the use of a physically accurate modeling technique, which leverages large G&G datasets and produces results which may be combined into an induced seismicity potential proxy. This proxy can be plotted to create a tool which is interpretable by the general public, geoscientists, and regulators alike. Various ISP proxies are explored, which can be simulated on complex regional fault systems constrained by geologic and geophysical datasets. These ISP maps

Geomechanical Modeling of Fault Systems 179 could be quickly recomputed as the interpreted fault system is updated by using both available or future seismic surveys and wellbores, as well as other sources of information. This model will be improved by incorporating more realistic fault geometries and imposing variable material properties and pore pressure which reflect strati-graphic and depth ranges. Additionally, imposing injection wells as input to the model will help understand the magnitude of induced events. Finally, allowing faults in the model to propagate and intersect with other discontinuities, a feature available in the MPM geomechanical simulator, may help understand the time dependency of these events. In addition to enhancing the model, different approaches to validating the model may be explored, especially examination of principle stress axes rotation throughout the study area and comparison to robust, publicly available data.

8.7

Conflict of Interest

The authors wish to confirm that there are no known conflicts of interest associated with this publication and there has been no financial support for this work that could have influenced its outcome. The entire study reported in this publication is the outcome of FracGeo internal R&D efforts. The results of the study can be obtained free of charge by the public, any private or public company, state or federal government agency, or non-profit organization by contacting directly the authors.

Acknowledgements The authors thank Arman Khodabakhshnejad for his help with the MPM simulation of the one fault case shown in Figure 8.6.

References 1. I. Thompson and R.G. Boutilier, Modeling and measuring the social license to operate: Fruits of a dialogue between theory and practice, http://socialicense.com/publications/Modelling%20and%20Measuring%20the%20 SLO.pdf (2011). 2. T.H. Darrah, A. Vengosh, R.B. Jackson, N.R. Warner, and R.J. Poreda, Noble Gases Identify the Mechanisms of Fugitive Gas Contamination

180

3.

4.

5.

6. 7. 8.

9.

10. 11.

12. 13. 14.

Hydraulic Fracturing and Well Stimulation in Drinking-water Wells Overlying the Marcellus and Barnett Shales. Proceedings of the National Academy of Sciences Proc Natl Acad Sci USA 111(39), 14076–4081 (2014). Y. Aimene and A. Ouenes, Geomechanical modeling of hydraulic fractures interacting with natural fractures — Validation with microseismic and tracer data from the Marcellus and Eagle Ford. Interpretation 3(3), SU71–SU88 (2015). DOI: 10.1190/ INT-2014-0274.1 A. Ouenes, N. Umholtz, and Y. Aimene, Using geomechanical modeling to quantify the impact of natural fractures on well performance and microseismicity: Application to the Wolfcamp, Permian Basin, Reagan County, Texas. Interpretation 4(2), SE1–SE15, (April 2016). DOI: 10.1190/INT-2015-0134.1 A. Ouenes, Y. Kiche, L. Ouhib, R. Smaoui, M. Paryani, S. Poludasu, A. Bachir, and D. Balogh, Efficient Development of Unconventional Reservoirs Using 3G Workflows – Breaking the Silos to Achieve True Integration with Multidisciplinary Software. First Break 34 (May 2016). C. Scholz, The Mechanics of Earthquakes and Faulting, Cambridge University Press, New York (2002). DOI: 10.1017/CBO9780511818516 G.C.P. King and M. Cocco, Fault interaction by elastic stress changes: new clues from earthquake sequences. Advances in Geophysics 44, 1–38 (2001). Y.E. Aimene and J.A. Nairn, Modeling Multiple Hydraulic Fractures Interacting with Natural Fractures Using the Material Point Method. Society of Petroleum Engineers, Canada (2014). DOI: 10.2118/167801-MSD Sulsky, Z. Chen, and H. L. Schreyer, A particle method for history-dependent materials. Computer Methods in Applied Mechanics and Engineering 118, 179–196 (1994). J.A. Nairn, Material Point Method calculations with explicit cracks. Computer Modeling in Engineering and Science 4, 649–663 (2003). N. Umholtz and A. Ouenes, Optimal Fracing Near Faults - Quantifying the Interaction Between Natural and Hydraulic Fractures Using Geomechanical Modeling, Society of Petroleum Engineers, Canada (2015). DOI: 10.2118/176932-MS A. Holland, Peliminary Fault Map of Oklahoma, http://www.ou.edu/content/ogs/data/fault.html (2015). USGS Earthquake Archives, ANSS ComCat, http://earthquake.usgs.gov/ earthquakes/search Oklahoma Corporation Commission (OCC). Oil and Gas Data Files, Accessed January 5, 2015. http://www.occeweb.com/og/ogdatafles2.htm

9 Correlating Pressure with Microseismic to Understand Fluid-Reservoir Interactions During Hydraulic Fracturing Debotyam Maity

*

Gas Technology Institute, Des Plaines, IL

Abstract Hydraulic fracturing of shales and other tight formations has gained tremendous prominence over the past decade or so, and is set to grow in importance with increasing global energy demands. In recent years, new methods, which develop upon older techniques, have been proposed for the interpretation of multi-stage hydraulic fracturing data. A major issue has been validating these methods for wider applicability within the industry, considering some limitations that these techniques pose as a result of certain inherent assumptions. In this paper we take up a recently proposed diagnostic technique and analyze completion data from multiple shale gas wells. We correlate the results with available microseismic monitoring data and validate the interpretations made using this approach. Using actual stage-wise production contributions information from production logs, we also highlight the limitations of the current approach when it comes to predicting overall stage-wise completion quality. Keywords: Hydraulic fracturing, pressure, microseismic, completion, diagnostics

9.1 Introduction Analyzing pressure data as a means to understand treatment behavior has been a part and parcel of the oil and gas industry for many decades now. Pressure data can help identify fracturing behavior, proppant transport issues, screen out situations, limited entry calculations, etc., to name just a Email: [email protected] Fred Aminzadeh (ed.) Hydraulic Fracturing and Well Stimulation, (181–198) © 2019 Scrivener Publishing LLC

181

182

Hydraulic Fracturing and Well Stimulation

few of its uses. However, with the increasing use of multi-stage hydraulic fracturing treatments, many of these traditional techniques no longer hold validity, and newer techniques are considered desirable for improved treatment diagnostics. In this paper we consider a recently proposed diagnostic technique, which builds upon a very popular real time fracturing analysis technique, and demonstrate the utility as well as pitfalls associated with using such techniques.

9.2 Method 9.2.1 Pressure Data Analysis Simple fracture models were proposed by Perkins and Kern [1] who suggested that the fracturing pressure at the wellbore should be a power function of treatment time (Eq. 9.1) with a large value of the exponent indicating better fluid containment within the developing fracture.

p(t)

te

(9.1)

This and other work by Nordgren [2] formed the basis for the real time completion analysis technique as proposed by Nolte and Smith [3]. They concluded that fracture propagation may follow one of four predefined modes based on the slope of net pressure plotted against time as shown in Figure 9.1. The basic power law equation guiding the propagation of hydraulic fractures is given as follows:

log (p−pclosure)

e × log(t)

(9.2)

Recent proposals include modifications to the original Nolte-Smith approach to take into account the intermittent nature of fracture propagation and the well-known phenomenon of natural and hydraulic fracture interaction during treatment in shale plays [4, 5]. Developing upon the power function defined by Eq. 9.1, and considering the reference time (ti) for fracture growth initiation, we have:

p−pi = C (t−ti) t ti

p eC t ti t

(9.3) e

(9.4)

Correlating Pressure with Microseismic

p e p pi t

t ti

183 (9.5)

Based on Nolte’s generalization [6] for the bounds applicable on Eq. 9.1 for non-Newtonian fluids, we expect the exponent to range close to 0.25. If the fracture is dilating, the exponent will have a value close to 1. If the growing hydraulically created fracture interacts with natural fracture swarms, we expect rapid loss of fracturing fluid into the natural fracture system with the pressure behavior to be similar to fast leak off scenario (Figure 9.1, Mode IV) provided the pressure is stable or increasing. At the same time, we understand that as soon as pressure drops due to such fluid loss, the fissures tend to close as it drops below the stress holding them open. However upon closure, the pressure starts to rise again and there is a corresponding opening of closed fissures. This opening-closing cycle should produce rapid fluctuations in the exponent as an indicator of natural fracture interaction during treatment as will be shown later. This modified Nolte-Smith approach with varying reference time has been detailed by Pirayesh and others [4]. We use the same approach to estimate time variant exponent ‘e’ during treatment and use the results in our analysis. The calculations involve evaluating the exponent and constant at each point of time (during treatment) and then integrating the results from the reference to the current time.

et

pn ct

E

C

dp / dt pi / tn ti pn

pi

tn ti 1 tn ti

1 tn ti

et

(9.6)

(9.7)

tn

et dt

(9.8)

ct dt

(9.9)

ti

tn

ti

184

Hydraulic Fracturing and Well Stimulation

Net pressure

III

II I IV

Time

Figure 9.1 Nolte-Smith Interpretation Guide (log Pnet vs. log T). Mode I indicates constrained height with unrestricted extensional growth; Mode II indicates Stable growth; Mode III indicates restricted extension or screenout conditions; Mode IV indicates unstable fracture height growth.

Next we can estimate BHP by using the E and C values and identify the error from the actual BHP value available either through downhole pressure instrumentation or surface pressure data corrected for downhole conditions.

BHPest = pn + C (tn−ti)E

(9.10)

ε = |BHJPest – BHPcalc.|

(9.11)

If the calculated error mismatch, ε, is higher than a predefined threshold, the initial reference point (ti) is shifted to the current evaluation time. This provides us with a varying value for exponent e for the entire treatment, and this can then be used for our diagnostic interpretations. As an example, Figure 9.2 shows pressure match obtained for a reference stage and the error in evaluation. We note that for this analysis, we used a threshold limit of 5 psi in order to obtain a close match between the estimated BHP and the calculated BHP. We note that it is preferable that we use downhole pressure data instead of calculating BHP from surface measurements. The potential pitfalls of using BHP calculated from surface data compared to downhole measurements will be looked into later on. After validation of adequate accuracy in the net pressure prediction, the exponent variability with completion time can be interpreted as required. Figure 9.3 shows the variability of exponent ‘e’ for the same treatment data subset shown earlier, and we can clearly identify those sections of the treatment where we either had Mode I, Mode III or Mode IV fracturing based

Correlating Pressure with Microseismic 1.202

x 10 4 Calc. BHP Modeled BHP

15 Δ Pressure (psi)

1.2

Pressure (psi)

185

1.198 1.196

10

5

1.194 64 66 68 Time (minutes)

(a)

0

62

(b)

66 68 64 Time (minutes)

Figure 9.2 (a) Comparison of calculated BHP and estimated BHP (Eq. 9.10) and (b) observed ΔP mismatch for a short window extracted from completion data for a sample hydraulic fracturing stage.

Exponent, e

1 0.5 0 -0.5 -1

62

63

64

65 66 Time (minutes)

67

68

69

Figure 9.3 Modeled modified Nolte-Smith exponent ‘e’. The region highlighted from e = 0.75 to 1 indicates restricted flow and fracture dilation, the region highlighted from e = 0.1 to 0.3 indicates unrestricted fracture extension and from e = -1 to -0.1 indicates natural fracture interactions or rapid height growth.

on the distribution of the exponent values. We observe that there is significant interaction with natural fractures and we also observe that towards the end of the period in question, extensional growth of the fracture network occurs with fluid injection. The exact range used for the zones indicating the three modes are: -1.00 to -0.10 for mode IV; 0.10 to 0.30 for Mode I and 0.75 to 1.00 for Mode III. While real time diagnostic application has already been studied for multi stage shale gas completions by Soliman [5], we want to look at the predicted fracturing behavior and identify possible ways to characterize the completion effectiveness using the modeled parameters. This is achieved

186

Hydraulic Fracturing and Well Stimulation

by defining certain diagnostic parameters computed for each stage. For the three modes (I, III and IV) identified, we calculate two separate diagnostic parameters as defined below: tend

DP1

S

tts

| e |Δt

(9.12)

Δt

(9.13)

t 1

tend

DP2

S

tts t 1

Here δ represents the Kronecker delta, tend represents the end of evaluation period, and tS is the set with the time stamps that correspond with the value of exponent ‘e’ falling within the three modes in question. Finally, Δt is the time step in seconds. The reason for using the above definitions for the diagnostic parameters DP1-S and DP2-S was that the extent that the exponent ‘e’ stays within each modal region should provide an indication of how much extensional growth, dilatational growth as well as interaction with natural fractures occurs during completion.

9.2.2 Microseismic Data Analysis Microseismic events occur due to stress changes and pressure variations as a result of fluid being injected into the formation during hydraulic fracturing [7]. These events occur due to small movements over very small rock area in either shear, tensile or complex mode failure [8]. In the 1940’s, Gutenberg and Richter identified a relationship between the frequency of earthquakes and their magnitudes, and this relation has been observed to be universal and hold for local and micro level seismicity as well [9]. The relation is defined as follows:

log10 N = 1−bM

(9.14)

Where M is the magnitude, N is the number of events within a particular magnitude/ time or spatial bin. Constants ‘a’ and ‘b’ are derived empirically from observed seismicity distribution. Figure 9.4 shows a sample b-value calculation based on 30 year data available from the Global CMT (Centroid Moment Tensor) catalog for the state of California (United States) highlighting the linear nature of the distribution outside the roll-off

N = No. earthquakes M

Correlating Pressure with Microseismic

102

187

b=1

100

6

7 8 Magnitude (M)

9

Figure 9.4 Gutenberg-Richter magnitude frequency relationship for earthquake data from California (CMT catalog).

regions and the expected slope (b-value) of 1 for tectonic earthquakes [10]. While evaluating b-value, care should be taken to ft the data using maximum likelihood method [11]. Datasets with very few earthquakes should not be considered for analysis. Also, error in magnitude can cause significant uncertainty in the slope estimation as well as identifying the MC (Magnitude of completeness). In our analysis, we assume that the errors in magnitude within our catalogs are negligible. For local seismicity scenarios, variations in b-value can be attributed to reservoir heterogeneity, thermal gradients across the perturbed zone, applied stress through injection as well as other factors [12]. For induced fracturing, this relationship was originally observed by Scholz [13] and has been further validated in recent years [14, 15]. Finally, recent studies have also indicated some correlation between observed b-value distribution and the local stress regime guiding rock failure [16, 17]. In general, when it comes to local or microseismicity, b values less than 1 are rare and represent compressive failure modes; b values close to one indicate strike-slip regime similar to observations during failure of faults (from earthquake seismology) and b values higher than 1 are indicative of extensional failure modes.

9.3 Data We apply this technique on two separate wells from the Appalachian Basin (Marcellus shale gas play). Both the wells were completed as essentially dry gas wells with reasonable water cuts. The choice was based on the fact that both these wells involved microseismic monitoring during completion,

188

Hydraulic Fracturing and Well Stimulation

which was a necessary element of our analysis. Well # 1, located in northern Pennsylvania, involved the pumping of ~ 300,000 lbs. of proppant for each stage with a total of 18 stages. Well # 2 located in western Pennsylvania again involved the pumping of ~ 300,000 lbs. of proppant for each individual stage with a total of 13 stages. Both completions were monitored by downhole sliding microseismic monitoring arrays placed in nearby horizontal observation wells. For b-value analysis, we used microseismic event catalogs generated by using the standard P (compressional) and S (shear) wave travel time inversion approach. Production logs were run post treatment for both the wells, and these are also used in our study. For Well # 1, we also have data from OBMI log, and we also use the interpreted stagewise fracture density from this log for our analysis.

9.4 Results As stated earlier, we apply both these techniques, namely modified NolteSmith approach, to fracture diagnostics as well as stage wise b-value mapping on the data from the two wells under study. For both wells, we generate the diagnostic parameters (DP1-S and DP2-S) for each of the fracturing modes under study defined by subscript S (Mode I, III and IV). Based on the identified parameter values, we cross-correlate the same with available production log data (fractional gas flow) from post completion production logging runs. Figure 9.5 shows how the modeled diagnostic parameter results for Well # 1 correlate with observed stage-wise productivity from production logs. We notice weak positive correlation for both mode I and mode IV results but no correlation with mode III results. Figure 9.6 shows the same parameters evaluated for Well # 2 under study and their correlation with stage-wise productivity. Once again, we observe weak positive correlation for both the mode I and mode IV parameters and no correlation whatsoever with mode III results. Since observations from both sets of analysis seem to validate each other, we can argue that this observation is non-unique and should hold for more completions. We do note that the correlations are weakly positive and may not signify much. However, in essence, these observations indicate that the productivity from any hydraulically fractured stage shows slight correlation with the extent of extensional fracture growth taking place during the treatment as well as any interaction with natural fractures observed during treatment. Moreover, we can clearly state that the degree of dilatational fracture growth has no impact on the productivity of the completed zone. These observations are also intuitive as we would

Correlating Pressure with Microseismic 140

600

120

500

100

400 DP 2-l

DP 1-l

80 60

189

300 200

40 20

100

y = 353.25x + 73.603 R2 = 0.3066

0 0 (a)

0.05 0.1 Fractional gas

0 0

0.15 (b)

1600

1800

1400

1600

1200

1400 DP 2-lII

DP 1-lII

800

800 600

400

400 y = 397.23x + 905.92 R2= 0.0032

200 0 0 (c)

0.05 0.1 Fractional gas

0.15

0.05

0.15

y = 478.99x + 989.48 R2 = 0.0041

200 0 0 (d)

0

0.05 0.1 Fractional gas

0.15

1400 0

0.1

y = −1178.8x − 318.41 R2 = 0.2997

1200 1000 DP 2-lV

−200 DP1-lV

0.15

1000

600

−300

800 600 400

−400

200

y = 2384.8x + 772.05 R2 = 0.3076

−500 0 0

−600 (e)

0.05 0.1 Fractional gas

1200

1000

−100

y = 1502.6x + 286.31 R2 = 0.3206

Fractional gas

(f)

0.05

0.1

0.15

Fractional gas

Figure 9.5 Production data (fractional gas flow) compared with extensional diagnostic parameters (a) (DP1-I), (b) DP2-I, dilatational diagnostic parameters (c) DP1-III, (d) DP2-III, natural fracture interaction diagnostic parameters (e) DP1-IV and (f) DP2-IV for Well # 1.

Hydraulic Fracturing and Well Stimulation

190

180

700

160

600

140

500 400

DP2-l

DP1-l

120 100 80

300

60

200

40 0 0

0

30 (b) 2000

1600 1400

1800 1600

1200

1400

1000

1200

DP2-lII

1800

800

600 400 y = 5.4387x + 1058.1 R2 = 0.0122

200 0 0 (c)

10 20 Fractional gas

0

30

10 20 Fractional gas

30

1000 0

10

20

30

900

y = −7.1587x −225.18 R2 = 0.3822

800 700 DP2-lV

DP1-lV

0

(d)

−150 −200 −250

600 500 400 300

−300

200

−350

y = 13.746x + 572.19 R2 = 0.4271

100

−400 (e)

y = 3.4788x + 1089.9 R2 = 0.0041

200

0

−450

30

800

400

−100

10 20 Fractional gas

1000

600

−50

y = 11.798x + 324.91 R2 = 0.4554

0

10 20 Fractional gas

(a)

DP2-lII

100

y = 2.9381x + 81.959 R2 = 0.4891

20

0 0 Fractional gas

(f)

10

20

30

Fractional gas

Figure 9.6 Production data (fractional gas flow) compared with extensional diagnostic parameters (a) (DP1-I), (b) DP2-I, dilatational diagnostic parameters (c) DP1-III, (d) DP2-III, natural fracture interaction diagnostic parameters (e) DP1-IV and (f) DP2-IV for Well # 2.

Correlating Pressure with Microseismic

191

not expect fractures ballooning due to fluid fill-up to have any impact on productivity as any far field fracturing due to stress perturbation may not be contributing due to them being spatially isolated. Extensional growth and growth into natural fractures, on the other hand, should lead to more productivity due to higher fractured area through the connected hydraulically created as well as natural fractures. Next we look at the b-value analysis results and how they correlate with the modeled parameters shared above. Careful selection was made from the microseismic event catalogs to make sure that the analysis was valid, including removing outliers. Those stages with very low event count were not considered in this analysis (stages 1 through 5 for well # 1). Finally, fitting to identify slope was done using maximum likelihood technique as mentioned earlier. We understand that higher b values (and consequently higher fractal dimensions) are indicative of more complex fractured network or values higher than 1, indicating extensional fracture growth [18]. In our analysis, for all stages studied, we found b-values close to or higher than 1. Any b-value close to or higher than 2 could be a result of microseismic data quality or could indicate fluid-rich completions. Also, as observed from Figures 9.7 and 9.8, b-value shows a much stronger correlation with extensional growth parameters compared to natural fracture interaction related parameters. Finally, b-value shows no correlation with dilatational fracture swelling related parameters. This is again expected as extensional growth through fracture propagation should create extensive three dimensional microseismicity and so should interaction with natural fractures. However, we note that with fluid flling up dilating fractures, seismicity will be limited in size (small shear tip or far field failures). Next, we look at correlation with OBMI log in Figure 9.9. As observed from Figure 9.9, we can clearly see no apparent correlation between the derived fracture density from OBMI logs and the modeled diagnostic parameters. This is expected since the OBMI logs provide a snapshot of wellbore or near wellbore fracturing and cannot estimate the presence or absence of actual fractures in the formation. Eyeballing the modeling results using the pressure data can provide indicators to make judgement calls by identifying stages with significant interactions with natural fractures during treatment. Figure 9.10 shows sample stages from Well # 1 showing the modeled exponent. We validate the observations by comparing the corresponding b-values as shown in Figure 9.11. As observed from results for the two stages shared in Figure 9.10, the stage corresponding to Figure 9.10a (stage ‘A’) shows a relatively lower

Hydraulic Fracturing and Well Stimulation 3.5

3.5

3

3

2.5

2.5 b-value

b-value

192

2 1.5

2 1.5

1

1

0.5 0 0

50

100

0 (b)

3

3

2.5

2.5 b-value

3.5

2 1.5

400

600

2 1.5

1

1 y = −0.0002x + 2.2792 R2 = 0.0248

0.5

y = −0.0002x + 2.2418 R2 = 0.0162

0.5

0

0 500

1500

b-value

(c)

1000 DP1-lII

(d) 3.5

3.5

3

3

2.5

2.5

2 1.5

y = −0.0026x + 1.1803 R2 = 0.2976

0

2000

b-value

0

−200

−400 DP1-lV

500

1000 DP2-lII

1500

2000

2 1.5

1

1

0.5

0.5

0 (e)

200 DP2-l

3.5

−600

y = 0.0028x + 1.0139 R2 = 0.5757

0

150

DP1-l

(a)

b-value

0.5

y = 0.0116x + 0.9782 R2 = 0.5601

y = −0.0014x + 1.2298 R2 = 0.2775

0

0

0

(f)

500 DP2-lV

1000

Figure 9.7 Microseismic derived b-value compared with extensional diagnostic parameters (a) (DP1-I), (b) DP2-I, dilatational diagnostic parameters (c) DP1-III, (d) DP2-III, natural fracture interaction diagnostic parameters (e) DP1-IV and (f) DP2-IV for Well # 1.

3

3

2.5

2.5

2

2 b-value

b-value

Correlating Pressure with Microseismic

1.5

1.5 1

1 0.5

0.5

y = 0.0067x + 1.348 R2 = 0.5404

y = 0.0016x + 1.3941 R2 = 0.5164

0

0 0

50

(a)

100 DP1-l

150

0

200

200

400 DP2-l

(b)

3

3

2.5

2.5

2

2 b-value

b-value

193

1.5

600

800

1.5 1

1 0.5

0.5

y = 0.0002x + 1.8431 R2 = 0.0568

0 0

500

(c)

1000 DP1-lII

1500

y = 0.0002x + 1.8588 R2 = 0.0574

0 0

2000

500

1000 DP2-lII

(d)

1500

2000

1000

3

900 2.5

800 700

b-value

b-value

2 1.5 1

−600 (e)

−400 DP1-lV

400 200 y = 364.95x −68.817 R2 = 0.4384

100 0

0 −200

500 300

0.5

y = −0.0022x + 1.421 R2 = 0.4581

600

0

0 (f)

1

2

3

DP2-lV

Figure 9.8 Microseismic derived b-value compared with extensional diagnostic parameters (a) (DP1-I), (b) DP2-I, dilatational diagnostic parameters (c) DP1-III, (d) DP2-III, natural fracture interaction diagnostic parameters (e) DP1-IV and (f) DP2-IV for Well # 2.

Hydraulic Fracturing and Well Stimulation

194

140

140

y = −0.2374x + 62.421 R2 = 0.0172

120

80

80

OBMI

100

OBMI

100

60

60

40

40

20

20

0

0 0

50

100

150

0

DP1-l

(a) 140

y = 0.0271x + 14.913 R2 = 0.0268

120

y = 0.0249x + 14.756 R2 = 0.0262

80

OBMI

80

OBMI

600

140 120

60

400 DP2-l

100

60

40

40

20

20

0

0 500

(c)

1000 1500 DP1-lII

OBMI

y = 0.0865x + 70.236 R2 = 0.025

2000

0 (d)

140

140

120

120

100

100

80

80

OBMI

0

60

−400

−200 DP1-lV

500

1000 1500 DP2-lII

2000

y = −0.0443x + 66.422 R2 = 0.0199

60

40

40

20

20 0

0 −600

200

(b)

100

(e)

y = −0.0631x + 63.648 R2 = 0.021

120

0

0 (f)

500 DP2-lV

1000

Figure 9.9 OBMI log derived fracture density compared with extensional diagnostic parameters (a) (DP1-I), (b) DP2-I, dilatational diagnostic parameters (c) DP1-III, (d) DP2-III, natural fracture interaction diagnostic parameters (e) DP1-IV and (f) DP2-IV for Well # 1.

Correlating Pressure with Microseismic

195

e

1 0

e

-1 20 (a) 1

25

30

35

40 45 Time (minutes)

50

55

60

65

0

-1

25

30

35

(b)

40 45 50 Time (minutes)

55

60

65

Figure 9.10 Modeled exponent ‘e’ for two stage treatments during proppant injection phase. Subplot (a) shows lower productivity for stage ‘A’ compared to subplot (b) showing productivity for stage ‘B’ from production log data.

Cumulative number

Cumulative number

103

102

101

100 (a)

b = 1.1 +/- 0.09 a = 0.8218 -3

-2.5

-2

-1.5 -1 Magnitude

-0.5

0

102

101

100 (b)

b = 1.49 +/- 0.1 a = 0.52611 -2.5

-2

-1.5 -1 Magnitude

-0.5

Figure 9.11 Results from b-value analysis for (a) stage ‘A’ and (b) stage ‘B’ using Maximum Likelihood approach.

degree of interactions with natural fractures. Moreover significant sections of the treatment show dilatational behavior compared to stage corresponding to Figure 9.10b (stage ‘B’). These observations are validated by actual observations from production log data with stage ‘A’ showing 1.135% gas flow contribution compared to 14.139% for stage ‘B’ from the post completion production log. Furthermore, the corresponding b-value evaluation shows significant differences with stage ‘B’ showing a higher value (1.5), indicating higher fracture complexity compared to stage ‘A’ (1.0).

196

Hydraulic Fracturing and Well Stimulation

9.4.1 Pitfalls in Analysis We need to carefully consider the limitations and assumptions made in our analysis and highlight adequate caution required when conducting similar analyses. Before looking into the models themselves, the most important constraint with any diagnostic methodology is the actual data quality collected from the field and used in the analysis. With pressure data, downhole pressure measurement is extremely rare in long multi-stage horizontal hydraulic fracturing programs due to cost issues. Downhole data is necessary, as an assumption of net pressure being equal to the calculated BHP can be highly flawed due to uncertainties in models used for these calculations, including corrections for air entrainment, frictional losses and impact of proppant loads. With microseismic data, errors in inversion, limitations of array design, deployment issues, etc. can significantly alter the overall quality of the microseismic catalogs and thereby undermine any interpretations. When it comes to the actual physical fracture model used, we need to highlight that it is by very nature highly simplistic. Moreover, the variability in exponent, though very useful, can be interpreted in multiple ways. As an example, negative exponent values could mean natural fractures as suggested in this work. But at the same time, they could also mean uncontrolled rapid height growth into lower closure stress zones (particularly with decreasing pressure) or interaction with local faults. Similarly, with b-value analysis, sometimes the data artifacts require manual mapping of the slopes resulting in significant non-uniqueness in the selected values. Therefore we believe that for proper utilization of this technique, more robust ways of analyzing data, as well as analysis of other information in addition to those shared in this work, may be necessary. All of these considerations will influence our future work with new wells.

9.5 Conclusions Novel completion diagnostic techniques such as those applied in this study provide valuable tools, which can be very useful in helping people understand the behavior of long lateral multi-stage hydraulically fractured wells. Judicious selection of data, analysis methodology and a careful consideration of potential pitfalls are also necessary in order to add value to any such diagnostic workfow. In this study, we have demonstrated two ways of identifying potential fracture growth mechanisms available today and have tried to correlate the two to highlight a reasonable match between the results as per our observations. However, a more careful analysis of

Correlating Pressure with Microseismic

197

methods as well as better models are deemed necessary before any such technique can find widespread use within the fracking industry. This is because of a high degree of uncertainty and the non-uniqueness possible in the interpretations.

9.6 Acknowledgements This work was supported by RPSEA project number 11122-20. I also acknowledge WPX Energy for providing access to active hydraulically fractured wells of opportunity in the Marcellus shale play, and Schlumberger for acquiring the production log data used in this study. I would also like to acknowledge valuable contributions and constructive suggestions by Iraj Salehi and Jordan Ciezobka from the Gas Technology Institute.

References 1. T.K. Perkins and L.R. Kern, Widths of hydraulic fractures. J. Petr. Tech. 13, 937–949 (1961). 2. R.P. Nordgren, Propagation of a vertical hydraulic fracture. Soc. Petrol. Eng. J. 253, 306– 314 (1972). 3. K.G. Nolte and M. Smith, Interpretation of fracturing pressures. J. Petr. Tech. 33, 1767– 1775 (1981). 4. E. Pirayesh, M.Y. Soliman, and M. Rafiee. Make decision on the fly: A new method to interpret pressure-time data during fracturing – application to frac pack. SPE Annual Technical Conference and Exhibition (2013). 5. M.Y. Soliman, M. Wigwe, A. Alzahabi, and E. Pirayesh, Analysis of fracturing Pressure data in heterogeneous shale formations. Hydraulic Fracturing J. 1(2), 8–13 (2014). 6. K.G. Nolte, Determination of fracture parameters from fracturing pressure decline, SPE Annual Technical Conference and Exhibition (1979). 7. N.R. Warpinski, S.L. Wolhart and C.A. Wright, Analysis and prediction of microseismicity induced by hydraulic fracturing. SPE J. 9(1), 24–33 (2004). 8. M.F. Kanninen and C.H. Popelar: Advanced Fracture Mechanics, pp. 563, Oxford University Press, New York (1985). 9. B. Gutenberg and C.F. Richter: Seismicity of the Earth and Associated Phenomena, pp. 16–25, Princeton University Press, Princeton, New Jersey (1949). 10. A.M. Dziewonski, T.A. Chou, and J.H. Woodhouse, Determination of earthquake source parameters from waveform data for studies of global and regional seismicity. J. Geophys. Res. 86, 2825–2852 (1981).

198

Hydraulic Fracturing and Well Stimulation

11. K. Aki, Maximum likelihood estimate of b in the formula log (N) = a - bM and its confidence limits. Bull. Earthq. Res. Inst. Tokyo Univ. 43, 237–239 (1965). 12. J. Farrell, S. Husen, and R.B. Smith, Earthquake swarm and b-value characterization of the Yellowstone volcano tectonic system. J. Volcanol. Geoth. Res. 188, 260–276 (2009). 13. C.H. Scholz, The frequency-magnitude relation of microfracturing in rock and its relation to earthquakes. Bull. Seismol. Soc. Am. 58, 399–415 (1968). 14. S. Wessels, M. Kratz, and A. De La Pena, Identifying fault identification during hydraulic simulation in the Barnett shale: Source mechanism, B values, and energy release analyses of microseismicity. 81st SEG Annual Meeting Expanded Abstracts 30, 1463–1467 (2011). 15. S. Sil, B. Bankhead, C. Zhou, and A. Sena, Analysis of B value from Barnett shale microseismic data, 74th EAGE Conference & Exhibition (2012). 16. M. Grob and M. van der Baan, Inferring in situ stress changes by statistical analysis of microseismic event characteristics. The Leading Edge 30(11), 1296–1302 (2011). 17. M. Grob and M. Van der Baan, Statistical analysis of microseismic event characteristics helps monitor in-situ stress changes, 21st Canadian Rock Mechanics Symposium (2012). 18. T. Urbancic, A. Baig, and S. Bowman, Utilizing B-values and fractal dimensions for characterizing fracture complexity, CSPG/ CSEG/ CWLS GeoConvention (2010).

10 Multigrid Fracture Stimulated Reservoir Volume Mapping Coupled with a Novel Mathematical Optimization Approach to Shale Reservoir Well and Fracture Design Ahmed Alzahabi1*, Noah Berlow2, M.Y. Soliman1 and Ghazi AlQahtani3 1

Petroleum Engineering Department, UH Energy Research Park (ERP), University of Houston, Houston 2 Department of Electrical and Computer Engineering, Texas Tech University, Lubbock, USA 3 Special Simulation Studies Unit Saudi Aramco

Abstract This paper introduces a grid-based fracture-stimulated reservoir volume (SRV) concept. SRV is defined as a volume occupied by fluid in a fracture, whether created or caused by intersection with natural fractures. Fracturing of the optimum zones is believed to contribute to higher hydrocarbon production from shale and tight formations. This requires choosing placement of fractures along the designed path of horizontal wells to maximize expected SRV. Additionally, a new linear programming-based approach to mathematically optimize the placement of SRV in shale reservoirs is presented. The approach may be useful in pad drilling and fracturing as well as development of applications for use with shale formations. This work aims to globally optimize the placement of surface well pads, the location and number of wells attached to the pads, and the location of the fractures throughout the wells. This optimum placement will also take into account numerous practical constraints, including the length of wells, the number of wells associated with a pad, numerous overlap constraints inherent in unconventional gas and oil well development, the spacing between wells and fractures, etc.

*Corresponding author: [email protected] Fred Aminzadeh (ed.) Hydraulic Fracturing and Well Stimulation, (199–225) © 2019 Scrivener Publishing LLC

199

200

Hydraulic Fracturing and Well Stimulation

One approach to this challenge to optimize is based on maximizing the fracturability index (FI) values assigned to the cells of the model values explored by the final network, and will be constrained by the previously mentioned considerations as well as a global maximum number of wells and a maximum development budget. In addition, the mathematical framework allows for easy extensibility to other constraints, and can be customized based on these constraints. Keywords: Stimulated reservoir volume, shale reservoir, optimization, well and fracture placement design, fracture network, well pad

10.1 Introduction Simulated reservoir volume (SRV) has a long history of use in defining the effect of fracturing in shales. Substantial evidence from sonic logs and production data from shale wells shows that certain segments of the wells make up 70% of the total production of wells. This paper presents a concept for identifying SRV in shale rock. Creating hydraulic fractures leads to fracture network growth. Fracture growth interaction with existing natural fractures causes complexity. Complexity is a resultant network of induced and existing fractures. SRV is used to account for this resultant complexity. These complex networks have a substantial impact on well performance in shale and tight rocks. The shape of SRV can be predicted from stimulated and shear propped fractures, while the volume can be correlated with fracture network length. Britt et al. [1] & Cipollaet et al. [2] discussed geomechanics of a shale prospective and fracture complexity. The size of the SRV is correlated to treatment volume based on microseismic measurement. Figure 10.1 shows the relationship between treatment volume and fracture network length for five vertical Barnett shale wells, modified after Fisher et al. [3, 4]. Mayerhofer et al. (2008) introduced the SRV concept as a 3D size of created fracture network, and defined SRV as a correlation parameter for well performance. Mayerhofer et al. [5] linked SRV with well performance of shale reservoirs. A direct relationship is demonstrable between fracture network length and SRV, as shown in Figure 10.1 for Barnett shale wells (modified after Fisher et al. [3]). Anderson et al. [6] defined SRV linked to the horizontal well by stimulated reservoir width, areal extent and fracture half-length. Zhou et al. [7] introduced a method for identifying anisotropic regions in unconventional hydrocarbon reservoirs. Anisotropy can be indicative of sweet spot zones for fracturing and for drilling a productive well. Seismic amplitude data from receivers along two orthogonal lines radiating from a seismic source

Fracture Stimulated Reservoir Volume Mapping

Fracture network length, ft.

30000 25000

201

Fracture network length = –4E–05 Vol.2 + 2.108 Vol. + + 1920.2 R2 = 0.9839

20000 15000 10000 5000 0 0

2000

4000

6000

8000 10000 12000 14000 16000 18000 20000 Fluid volume bbl

Figure 10.1 Fracture network length as a function of fluid volume injected for five vertical wells of Barnett (modified after Fisher et al. [3]).

is used. Sil et al. [8] introduced a method to calculate fracture parameters from common well log data. Fracture parameters can indicate sweet spot zones in unconventional rock. Microseismic mapping is currently used to map SRV in shale rocks. It is also used as a tool to diagnose the effectiveness of the created hydraulic fractures, especially in multistage fracturing in horizontal wells. Zhang et al. [9] introduced the SRV equation as follows: n

SRV

SRA H f

Ai Hp ,

(10.1)

i 1

where Hf is the fracture height; SRA is the stimulated reservoir area; Hp and is the reservoir thickness. One limitation is implicit: it seems to be only valid when hp = hf. Cheng et al. [10] presented an SRV formula as a function of average fracture length, average height of fractures, maximum number of fractures, maximum horizontal stress and minimum horizontal stress as follows: H

nc 2 i i

SRV

h

4 x h 1

2

1

H h

(10.2)

202

Hydraulic Fracturing and Well Stimulation

where xi = Average length of the fractures in the ith cluster hi = Average height of the fractures in the ith cluster, m nc = Number of clusters σH = Maximum horizontal in situ stress, N/m2 σh = Minimum horizontal in situ stress, N/m2 Additionally, a linear programming-based approach to mathematically optimize SRV is presented. This newly developed optimization approach improves the placement of fractures in quantifiably better zones in shale reservoirs, to guarantee optimality of the reservoir development plan given the available data and modeling constraints. This approach will be useful in pad drilling and development of applications applied to shale formations. This work will lead to the global optimization of the placement of surface pads, location and design of wells attached to the pads, and location of the fractures (SRV) throughout the wells. This design will also take into account other practical design constraints, including length of wells, number of wells associated with a pad, numerous overlap constraints inherent in unconventional gas and oil well development, etc. The development will be optimized based on maximization of the FI values explored by the final network, and will be constrained by the previously mentioned considerations, as well as a global maximum number of wells and a maximum development budget. In addition, the mathematical framework allows for easy extensibility to other constraints, and can be customized based on the problem constraints. Fracture parameters such as half-length, azimuth, width and height can be used extensively in the fracture modeling process. Fracture geometry can be modeled instead through SRV, as an estimated fracture volume can give a better description of fracture parameters. SRV is represented by grid like geometry that has a value of FI greater than the predefined threshold of 0.5. The SRV consists of a group of cells as shown in Figure 10.2, and SRV varies from one stage to the next. The next section details the geometric interpretation of the SRV representation, that is, a group of grids characterized by high values of FI. Before entering into discussion of the mathematical definition, it is necessary to set the objective of developing a method to predict SRV location and number using an input map of FI’s. Such a method, when coupled with mathematically developed code, could help in exploitation of the shale resource with the minimum number of wells and minimum number of fracture stages. The paper aims at: • Planning and automating an optimum well path and optimum fracture design in shale and tight formation;

Fracture Stimulated Reservoir Volume Mapping

203

Figure 10.2 Multigrid based SRV.

• Establishing a process of choice of maximum SRV for future initiation of fractures; • Finding out the optimum number of fracture stages; • Optimizing the number of SRV’s. Previous work [11] identified an index to help prioritize fracture position and scheduling. Mathematical optimization using Integer Programming (IP) proved its superior performance in vertical well placement (for details on its performance, see AlQahtani et al. [12]). Computational concepts such as dynamic programming and graph theory may be useful in exploration of algorithms applied to a wide range of oil and gas optimization topics, the most important part of which is the computational methods used to solve them, so that an optimum placement can be obtained. Since the problem is a mathematically based method, the problem definition is first outlined in the next section.

10.2 Problem Definition and Modeling 10.2.1 10.2.1.1

Geometric Interpretation Fracture Geometry

Hydraulic fracture geometry dimensions may be calculated using analytical approaches based on net pressure, fluid and rock properties. Another

204

Hydraulic Fracturing and Well Stimulation

common approach is microseismic monitoring, which fits a rectangular box to the microseismic event locations along the horizontal well path. SRV can be estimated based on the volume of the rectangular or principal component box, or by summing a series of volumetric boxes (e.g., Mayerhofer et al. (2008)). For a realistic approach, SRV is used as a representation of fractures connected to the wells. A stimulated approximation of grids is used to represent the hydraulic fracture.

10.2.2

The Developed Model Flow Chart

In this paper an approach for placing surface well pads and fractures in shale rock is shown in Figure 10.3. The solution is presented in Figure 10.3 and Figure 10.4 as follows. Figure 10.5a and b illustrate the two allowable designs for placing fractures and then SRV’s.

10.2.3

Well and Fracture Design Vector Components

1. 2. 3. 4. 5. 6. 7. 8. 9.

Number of transverse fracture stages per well (5–50). Number of wells of single pad (5–40). Number of perforation per stage (1–6). Length of horizontal well (6,000–10,000 ft). Half Length of fracture (200–600 ft). Spacing (wells, fractures), (500–1,600 ft). Pay zone thickness, (200–1,000 ft). Reservoir boundary dimensions (Ye, Xe), (rectangular shape). Variable Stimulated Reservoir Volume (VSRV) with different variable conductivity. 10. The formation is heterogeneous. 11. The transverse fractures fully penetrate the majority of formation, except 10 ft. from the boundary. Fractures are contained within the formation. 12. Multiple transverse fractures are not identical (varied fracture conductivity, and varied fracture propped characteristics like length and network width).

10.3 Development of a New Mathematical Model In this section, a description of the mathematical model used to solve the above-mentioned problem is given, consisting of an introduction of

Fracture Stimulated Reservoir Volume Mapping Direct global solution

Load reservoir parameters and data

Generate FI values for reservoir

Collect user designed optimization parameters User selects desired pad locations Identify viable well pad locations

Preprocess: generate frac SRV for all wells in well pad reach Preprocess: generate well pad, well, frac LP model

Store model

All models built?

Join all well pad models via overlapping constraints

Solve global model via MLP solver Save solved global model and export for simulator

End

Figure 10.3 First recommended flowchart to be used as a utility in the optimization process; it shows the interface between pad design and SRV’s.

205

206

Hydraulic Fracturing and Well Stimulation Local pad –> global merging

Load reservoir parameters and data

Generate FI values for reservoir Collect user-designed optimization parameters User selects desired pad locations Identify viable well pad locations Preprocess: generate frac SRV for all wells in well pad reach Preprocess: generate well pad, well, frac LP model Solve well pad model via MLP solver Store model solution

All models solved?

Convert models from LP to link-based models Merge and trim models Save solved global model and export for simulator End

Figure 10.4 Second recommended approach of optimization process.

Fracture Stimulated Reservoir Volume Mapping

(a)

207

(b)

Figure 10.5 Four different SRV’s are located, whether staggered, or staggered and overlapping designs. (a) Staggered design. (b) Staggered and overlapping design.

methodology, Objective Function, the essential sets, variables, and constant parameters followed by presentation of the optimization procedure.

10.3.1

Methodology

The following describes the numerical formulation of the newly developed technology.

10.3.2

Objective Function

The problem can be formulated as follows:

max

Amn Xmn

FkYk

(10.3)

An alternative objective function is to maximize the net income obtained from unconventional reservoirs. The net income is calculated as the difference between total income from total hydrocarbons produced and the total capital and operating expenses, including optimum wells and fractures.

10.3.3

Assumptions and Constraints Considered in the Mathematical Model

Sets, variables and decision variables are assumed as follows. For detailed explanation, see geometric interpretation of parameters in Appendix C. where,

208

Hydraulic Fracturing and Well Stimulation

10.3.3.1

Sets

1. Amn, total FI values unlocked by the stage from node m to node n 2. Fk, total FI values unlocked by fracturing at node k 3. Lmn, length of stage from node m to node n 4. EI(n), edges inbound to node n 5. EO(n), edges outbound from node n 6. S(n), set of stages that connect to or pass through node n, represented as edges in the network 7. Ψ(n), set of starting nodes mutually exclusive with w 8. E, set of valid edges for the network 9. Λ(m, n), set of fractures accessible from the stage between node m and node n 10. Φ(n), set of nodes whose fracture SRV would intersect the fracture SRV of node n 11. Ω(n), set of stages intersected or interfered with by fracturing node n

10.3.3.2

Variables

1. Pi,j,k, total FI values unlocked by fracturing or placing well at node i, j, and k

10.3.3.3

Decision Variables m i , j ,k

2. w binary variable equal to 1 if well m goes through node i,j,k (m = 1…maxwell) 3. fim, j ,,nk binary variable equal to 1 if fracture n, extending from well m, goes through node i,j,k (n = 1… maxfrac) 4. Wn, well origin for single well originating at node n, continuous 5. Xmn, flow of connections from node m to node n, continuous 6. Υn, fracturing of node n, continuous 7. Snm, usage of stage connecting node m to node n, binary

10.3.3.4

Extended Sets

8. ψw (m,i,j,k), set of nodes that cannot have well if wim, j ,k 1 mn 9. ψf (m,n,i,j,k), set of nodes that cannot have well fi , j ,k 1

(10.4) (10.5)

Fracture Stimulated Reservoir Volume Mapping

209

10. Ωw(m,i,j,k), set of nodes that cannot have well if due to constraints on well orientation angle (10.6) 11. Ωf (i,j,k), set of nodes neighboring node (i,j,k) along the plane of minimum stress

10.3.3.5

Constant Parameters

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12.

MinSpace: Minimum spacing between fractures MinStageSpace: Minimum spacing between stages MaxStageSpace: Maximum spacing between stages MaxLength: Maximum allowed length for a single well MaxStages: Maximum number of active stages in a well MaxFPS: Maximum number of fractures per stage MaxWells: Maximum number of wells per pad MaxPads: Maximum number of pads per reservoir MaxPadCost: Maximum cost for global development MaxCost: Maximum cost for global development Thickness: Reservoir thickness WC: Cost for initial construction of a well originating from a pad 13. PCmn: Cost for construction of a well segment from node m to n

10.3.3.6

Constraints

The problem is subject to the following constraints: 1. Fracture half-length length (Xf) < 0.9 (D1 + D2) (10.7) 2. The fracture propagates only along the predefined direction of maximum stress at each location (10.8) 3. Fracture half-length (Xf) < Xe & Xf 0.5. (b) One SRV after applying the filter of FI > 0.5.

10.4.1

Simulation Model of Well Pad and SRV’s Evaluation

To build a fast and accurate mathematically optimized approach to locate wells and SRV’s in the shale reservoir model, a model with random natural fractures distribution to represent the complexity of shale rock is built. Figure 10.11 shows a comparison between optimized placements of fracks in terms of SRV versus uniform optimized cases; the parameters of the reservoir model are listed in Table 10.1. Number of fracks vs. cum. oil prod.

Cum oil prod., BBls thousands

140 Optimized

120

Uniform distribution

100 80 60 40 20 0

0

2

4

6

8

10

12

14

Number of fracks

Figure 10.11 Optimized placement of the same size SRV versus uniform distribution of SRV along one horizontal well.

Fracture Stimulated Reservoir Volume Mapping

215

Table 10.1 Parameters of the reservoir model used in validating the developed model. Well type

Horizontal wells

Reservoir dimension, ft.

10,000 × 10,000

Fracture half length, ft.

500, 600, 650, 700, 750

Number of hydraulic fractures/well

5–50

Number of wells/pad

2–40

Length of horizontal well

(6,000–10,000 ft.)

Half length of fracture, SRV half length

(200–600 ft.)

Spacing (wells, fractures)

(500–1,600 ft.)

Stage width

200 ft.

1. Input Parameters i. 3D FI excel sheet heat map ii. Min Fracture stage (SRV) spacing iii. Reservoir dimensions X, Y, Z iv. Number of wells: 2 wells (budget constraints), assuming we have two pads at the surface, each pad has 2 wells v. Horizontal well length 6,000 ft vi. Overlap constraints vii. Spacing between wells and neighboring fractures 2. Output Parameters viii. Number of fractures stages (SRV) per well and per the pad ix. The location of surface well pads per the model. x. The locations Fracture stages (SRV) along the horizontal wells xi. Optimum scheduling of fracture stages (numbered rank based on sum of FI values per the stage). xii. Well spacing xiii. Fracture dimensions (length & width assuming fracture propagates as a network through the whole fracture stage)

216

Hydraulic Fracturing and Well Stimulation xiv. xv.

Predict SRV’s location and number using an input map of FI’s. An optimum well path and optimum fracture design placement in shale and tight formations

Assumption • Symmetric planar bidirectional propagation of the fracture from the wellbore occurs perpendicular to the direction of minimum principal stress. • Wells are in the direction of the minimum horizontal stress (predefined by the user).

10.5 Results and Discussions This section is to evaluate the performance of FI via reservoir simulation. The Permian basin evaluated consists of hundreds of fractures per well. Table 10.2 shows the data ranges used to develop the model. As the optimization technique has evolved, statistical algorithms and many software packages have been developed to improve the understanding of fracture stages and the SRV concept, and complex simulations are being implemented to take into account a greater number of variation in input parameters; however, the importance of considering the power of FI correlation and its predefined cut-offs remains paramount. The model presented in this work is based on coupling the sweet spot proxy and optimization tool. It requires input maps of calculated Fracturability Indices. Contrary to detailed microseismic-based techniques, it requires downhole sensing tools.

10.6 Conclusions and Recommendations In this paper an analysis of placing optimum wells and optimum fractures in the sweet spot regions of shale reservoir is presented. These sweet spots are known as grid assigned high values of FI. The time and number of required stages to reach the objective function was investigated. The fracture spacing and well spacing was assumed to be non-even. SRV size is not equal among stages due to the heterogeneous nature of the rock

Fracture Stimulated Reservoir Volume Mapping

217

Table 10.2 Properties of the reservoir model used in validating the developed model. Mineralogical properties Parameters

Minimum value

Maximum value

Quartz, wt. %

6.00

75.0

Calcite wt. %

0.00

84.0

Clay wt. %

3.00

49.0

Pyrite wt. %

0.00

8.00

Petrophysical properties Photoelectric Index (Pe), barns/electron

2.61

5.71

Density ρz, g/cc

2.41

2.71

Geomechanical properties E, psi

0.38 E6

9.75 E6

ν, ratio

0.02

0.38

Reservoir properties Initial reservoir pressure, psia Thickness, ft. Model dimensions Porosity, % Permeability, Nano-darcy

5,330 900 80 × 80 × 5 Avg. : 9 Avg. : 208

represented in varied FI values. The advantage of this helps in reducing the cost of placing wells and fracture stages. The concluding remarks from this paper are listed below. In this work, a mathematical optimization approach for the placement of horizontal wells and hydraulic fractures within shale reservoirs was developed. The approach is able to provide a design that gives the optimal predicted stimulation of sweet spot locations that are identified by the use of the Fracturability Index. The technique suggests the optimal number of wells and fractures needed in order to drain the shale reservoir by

218

Hydraulic Fracturing and Well Stimulation

achieving maximum contact area, while respecting the physical and economic constraints. i.

ii.

A model that includes the coupling of geomechanical and mathematical optimization was determined for the well data by use of a sophisticated Integer Programming approach. It is believed that the proposed model arrived at in this analysis is the best of its kind in the industry. A comparison of our proposed model versus published models (although published models are based on other non-optimal algorithms), shows better results in terms of accuracy in placing fractures. As a final recommendation, more refined models could be proposed in future work involving the collection of more data. Geometric placement of SRiV’s and hydraulic fracture stages in shale and tight formations should be replaced by coupled approaches of sweet spot indices and optimization methodologies.

References 1. Britt, L.K. and Schoeffer, J., The geomechanics of a shale play: What makes a shale prospective? 2008. SPE-125525. 2. Cipolla, C.L., Warpinski, N.R., Mayerhofer, M.J., Lolon, E.P., Vincent, M.C., The relationship between fracture complexity, reservoir properties, and fracture treatment design. Presented at the 2008 SPE Annual Technical Conference held in Denver, Colorado, USA, 21–24 September, 2008, SPE-115769, doi: 10.2118/115769-MS. 3. Fisher, M.K., Wright, C.A., Davidson, B.M., Goodwin, A.K., Fielder, E.O., Buckler, W.S., Steinsberger, N.P., Integration fracture mapping technologies to optimize stimulations in the Barnett Shale. Presented at the SPE Annual Technical Conference and Exhibition held in San Antonio, Texas, 2002, SPE-77441. 4. Fisher, M.K., Heinze, J.P., Harris, C.D., Davidson, B.M., Wright, C.A., Dunn, K.P., Optimizing horizontal completions techniques in the Barnett Shale using microseismic fracture mapping. Paper 90051 presented at the SPE Annual Technical Conference and Exhibition, Houston, 26–29 September, 2004, doi: 10.2118/90051-MS. 5. Mayerhofer, M.J., Lolon, E.P., Warpinski, N.R. et al., What is stimulated reservoir volume? Presented at SPE Shale Gas Production Conference, Fort

Fracture Stimulated Reservoir Volume Mapping

6.

7. 8.

9.

10.

11. 12.

13.

219

Worth, Texas, Society of Petroleum Engineers, 2010, SPE-119890, doi: 10.2118/119890-MS. Anderson, D.M., Nobakht, M., Moghadam, S., Mattar, L., Analysis of production data from fractured shale gas wells. Presented at the SPE Unconventional Gas Conference held in Pittsburgh, Pennsylvania, 2010, SPE-131787, doi: 10.2118/131787-MS. C. Zhou and S. Sil, Fracture identifcation from azimuthal migrated seismic data. Patent US20130201795, 2013. S. Sil, R. Keys, R. Baishali, D. Foster, Methods for seismic fracture parameter estimation and gas flled fracture identifcation from vertical well log data. Patent EP2506039A3, 2013. Zhang, S.C., Lei, X., Zhou, Y.S., Xu, G.Q., Numerical simulation of hydraulic fracture propagation in tight reservoirs by volumetric fracturing. Pet. Sci., 12, 674–682, 2015, doi: 10.1007/s12182-015-0055-4. Cheng, Y., Guo, B., Wei, N., Prediction of fracture population and stimulated reservoir volume in shale gas/oil reservoirs. Presented at the SPE Asia Unconventional Resources Conference and Exhibition held in Australia, 2015, SPE-176833, doi: 10.2118/176833-MS. Alzahabi, A., Soliman, M., Berlow, N., Spinner, T., 2016. AlQahtani, G., Alzahabi, A., Kozyreff, E., Farias, I., Soliman, M., A comparison between evolutionary metaheuristics and mathematical optimization to solve the wells placement problem. Adv. Chem. Eng. Sci., 3, 4A, 30–36, 2013, doi: 10.4236/aces.2013.34A1005. Alzahabi, A., Soliman, M.Y., AlQahtani, G.D., Bateman, R.M., Asquith, G., Fracturability index maps for fracture placement in shale plays. Hydraul. Fract. J., 2, 1, 8–18, 2015. ISSN 2373-8197.

220

Hydraulic Fracturing and Well Stimulation

Appendix A Abbreviations E v FCD K Δσh FI E E E v ρ x(i,j) X Y BHP Qg cf Lf Pi L ΔX,ΔY Pnet Xe, Ye W F Dmin

Young’s modulus Poisons’ ratio Dimensionless conductivity Permeability Difference between minimum and maximum horizontal stress Fracturability Index Young’s Modulus, psi Plane Strain Modulus, psi Normalized Plane Strain Modulus Poisson’s ratio Density, lb./ft3 X_Y location of a fracture in the reservoir represents the location (I, j) in the shale formation grid Coordinate axis along well path, ft. Coordinate axis along fracture path, ft. Bottomhole pressure, psi Gas fow rate, Mscf/d Formation compressibility, psi-1 Fracture half length, ft. Initial reservoir pressure, psi Well Lateral length, ft. Model grid dimensions in x and y direction, ft. Net pressure, psi Rectangular Reservoir shape dimensions in x and y direction, ft. Horizontal well Fracture stage Min well spacing, ft.

Appendix B Definition of the Fracturability Index Used in the Well Placement Process In this model every cell is assigned a number based on the calculated FI using geomechanical properties assigned to each cell. The FI consists of a

Fracture Stimulated Reservoir Volume Mapping

221

range between 0 and 1. The objective function is achieved through a sum of the total FI values unlocked by the stage from node m to node n and the total FI values unlocked by fracturing at node k.

Appendix C Geometric Interpretation of Parameters Used in Building the Model Figure A10.12 demonstrates 5 wells attached to one pad. Figure A10.13 shows one SRV ideal shape and ideal geometry. Figures A10.14 through A10.20 list all geometric interpretation of network connections in wells and fractures. Figure A10.14 shows a single well designated by originating at node moving in the direction of minimum horizontal stress, where shows continuous well connections from node m to n. Figure A10.15 shows a Pad

D1

D2

Figure A10.12 Combination of 5 wells originating from one well pad.

222

Hydraulic Fracturing and Well Stimulation σ max

Frac network height, h

σ min

Frac network width, w

Frac stage half length, Xf

Figure A10.13 Stimulated reservoir volume (SRV) for one fracture stage.

Well origin region (Pad reach)

Pad location at surface Variable distance e.x. space of 3, next node 4 cells away

Angular movement of 1 (max allowed)

Well origin Xmn

Wn

(n) Straight, level well path. Only connection that allows for fracking

Figure A10.14 Basic network connections in a well.

continuous flow of nodes representing fracture origin. Figure A10.16 defines the bounds of each individual valid SRV. Figure A10.17 differentiates valid versus invalid fracture locations considering minimum fracture distance. Figures A10.17 and A10.18 describe the possible geometrical representation of the well path considering other existing SRV’s. Figure A10.19 introduces the main dimensions of one possible SRV including fracture stage height and width. Figure A10.20 explains how SRV is represented in our model, whereas discretized SRV is an approximation of SRV ellipse 3D, as represented in Figure A10.13.

Fracture Stimulated Reservoir Volume Mapping Yk = 0 Yk = 1

Node n

Node m

Po tential fracture locations (^(m, n)) Figure A10.15 Hydraulic Fracture stage nodes representation.

Min frac distance

Invalid frac (too close to other frac)

Figure A10.16 Valid versus invalid nodes for fracture stage.

223

224

Hydraulic Fracturing and Well Stimulation Frac height Frac width

Invalid frac* (intersects SRV)

Invalid frac* (intersects SRV)

*Handled by Φ(n)

Figure A10.17 Three-dimensional representation of fracture intersection constraints.

Valid well path!

Invalid well path* (within 50 ft of frac)

Frac height Invalid well path* (intersects SRV) Frac width Invalid frac (intersects SRV)

Invalid well path* (intersects SRV)

*Handled by Ω(n)

Figure A10.18 Valid paths for neighboring wells.

Fracture height Frac center, well node

Fracture network width/ fracture stage length

Figure A10.19 Fracture stage parameters.

Fracture Stimulated Reservoir Volume Mapping

225

σmin σmax

Discretized SRV (sampled as approximated ellipse centered at frac with height and width determined by frac type) Frac center

Frac half length

Approximate 50 ft gap (will need to be adjusted based on scale)

Figure A10.20 Discretized SRV for one fracture stage.

11 A Semi-Analytical Model for Predicting Productivity of Refractured Oil Wells with Uniformly Distributed Radial Fractures Xiao Cai1*, Boyun Guo1 and Gao Li2 1

University of Louisiana at Lafayette 2 Southwest Petroleum University

Abstract Refracturing is a cost effective method enhancing the sustainability of both depleted and good quality reservoirs. It is generally believed that refracturing technology does not always work in all reservoir conditions. It is highly desirable to have a quantitative method for predicting well productivity to select refracturing well candidates. Assuming radial fractures around wellbore are uniformly distributed, a semi-analytical model was derived in this study to predict productivity of refractured oil wells under pseudo steady state flow conditions. Result of a field case study indicates that the mathematical model over-estimates well productivity by about 10%. Sensitivity analysis with the mathematical model shows that the productivity of refractured wells increases non-linearly with the number of radial fractures around the wellbore. Sensitivity analysis with the mathematical model also shows that the productivity of refractured wells increases non-linearly with fracture conductivity, but the benefit of increasing fracture conductivity levels out beyond 2000 md-inch. Therefore, there is no need to create very high-conductivity fractures in refracturing operations. This paper provides petroleum engineers a handy tool to assess well candidates for refracturing. Keywords: Refracturing, oil wells, productivity, analytical model

*Corresponding author: [email protected] Fred Aminzadeh (ed.) Hydraulic Fracturing and Well Stimulation, (227–241) © 2019 Scrivener Publishing LLC

227

228

11.1

Hydraulic Fracturing and Well Stimulation

Introduction

Seeking sustainable methods of fossil fuels and addressing energy sustainability are major contents of sustainable energy engineering. Refracturing is considered one of the important technologies used to strengthen the sustainability of the oil and gas reservoirs. It is a hydraulic fracturing operation conducted on a well which has already been hydraulically fractured. Refracturing can sustainably increase the production rate and enhance ultimate oil/gas recovery from not only the depleted reservoir, but also the good quality reservoir. In the down turn of the oil and gas industry, companies are realizing that refracturing old wells is a more budget-friendly way to achieve cash flow than drilling and completing new wells [1, 2], so it is also a sustainable method used to increase the cost effciency of oil companies. While some people are certain that it works quite well, others are still unsure of the success rate of refracturing due to variation of reservoir conditions [3]. It is highly desirable to have a quantitative method for assessing reservoir conditions to select wells for refracturing. For previously fractured wells, it is possible to create a secondary fracture that is perpendicular to the first. The secondary orthogonal fracture can be created within a certain time window that, in turn, depends on reservoir properties [4]. The optimum time window for refracturing is in the order of months to years. Roussel and Sharma [5] presented a systematic methodology for selecting candidate wells for refracturing using production data. It considers the stress reorientation occurring around a fractured well, causing the refracture to propagate orthogonally to the initial fracture in underdepleted sections of the reservoir. Ren et al. [6] presented a novel method to determine the refracturing strategy and priorities of the candidate wells by comprehensive analysis, which considers reservoir properties, remaining oil distributions, and induced stress fields. According to Miller et al. [7], reservoir depletion from parent wells can create localized “pressure sinks.” When a nearby infill well is fractured, the hydraulic fracture grows towards areas of lowest stress (the depleted area) which can damage the parent well. Subsequently, the production potential of the new well is decreased because it is now draining from the same reservoir area as the parent well. With the advent of modern diversion-based refracturing techniques, it is possible to create a “pressure barrier” around the parent well that helps redirect the new fracturing treatment into other, virgin areas of the reservoir. By refracturing the parent well first, this may not only preserve the original production of

Productivity of Refractured Oil Wells 229 the parent well, but it may also increase the production of the parent well, as well as the new infill wells. Although refracturing adds to the capital expenditure of the project, project economics may be more favorable with an investment in parent well refracturing. A gap exists between reservoir data and well refracturing design. Particularly, a simple and accurate mathematical model is needed for quantitatively predicting the productivity of refractured oil wells. This study fills the gap. No literature has been found to disclose any method for predicting productivity of refractured wells where multiple radial fractures are created. Methods for predicting productivity of a well with a single planar fracture of infinite and finite conductivity are well documented in the literature [8–11]. The concept of matrix-fracture cross flow presented by Guo and Schechter [10] for single-fractured wellbore is employed in this study to develop a semi-analytical model for quantitatively predicting productivity of refractured wells.

11.2 Mathematical Model It is expected that multiple radial fractures are created in refracturing the same pay zone. Initiation and extension of these fractures can be simulated using commercial software [5]. Assuming uniformly distributed radial fractures (Figure 11.1), an analytical model was derived in this study to investigate the long-term productivity of refractured oil wells. Derivation

Radial Fractures

k= kXky hf

Lf

Figure 11.1 Sketch of radial fractures around a refractured vertical well.

230

Hydraulic Fracturing and Well Stimulation

of the analytical model is given in Appendix A. The resultant well productivity model is summarized in this section. The well production rate in Darcy’s units is expressed by Eq. (11.1). The derivation of Eq. (11.1) is shown in the Appendix A. Eq. (11.1) is created based on the Eq. (A11.18) in Appendix A for the purpose of practical application.

4nkh f B0 tan( )

Q

Lf

rw

Pd dx x

(11.1)

kx k y is where Q is liquid production rate, n is number of fractures, k the geometrical mean of horizontal permeabilities, hf is fracture height, μ is liquid viscosity, Bo is the formation volume factor of oil, a = π/n is the half angle between fractures, rw is the wellbore radius, Lf is fracture length, and pd is expressed as

pd

C1 x J1 (2i c x ) C2 x N1 (2i c x )

(11.2)

where J1 is a Bessel Function of the first kind in the order of 1 and N1 is a Bessel Function of the second kind in the order of 1 and

4k wk f (tan )

(11.3)

C1

pdT12 T02T11 T12T01

(11.4)

C2

pdT11 T02T11 T12T01

(11.5)

T01

rw J1 (2i c rw )

(11.6)

T02

rw N1 (2i c rw )

(11.7)

c

Productivity of Refractured Oil Wells 231

T11

L f J1 (2i c L f )

(11.8)

T12

L f N1 (2i c L f )

(11.9)

pd

pe

pw

(11.10)

where pe and pw are reservoir pressure and wellbore pressure, respectively. For any given data set, the integral in Eq. (11.1) can be evaluated numerically. The numerical integration procedure has been coded in an MS Excel spreadsheet which is available from the authors upon request. This spreadsheet was utilized in the field case study and sensitivity analyses that are presented in the following sections. Because the mathematical model involves analytical solutions for pressure drawdown determination and numerical integration for flow rate calculation, it is called a semi-analytical model.

11.3 Model Verification No clean data set is found for refractured wells to verify the reliability of the mathematical model. A real well completed with high energy gas fracturing was employed to check the model accuracy. It is the well PIC-25 in the Pirital Field located in Maturin, Venezuela. The well was tested in Cretaceous and Naricual-5 formations of super-low permeability [12]. Propellant-assisted perforating was performed in 6 intervals of 9 feet long each in the Naricual-5 formation. Test-derived and estimated data are shown in Table 11.1. The well was perforated in a 2,500 psi underbalance condition by spotting diesel in the tubing prior to perforating. A stabilized oil rate of about 100 BOPD was achieved in a 20 ft perforated interval. The model-predicted interval production rate is 113 stb/day, which is an over-prediction by 13%.

11.4 Sensitivity Analysis The reliability of the mathematical model depends on the accuracy of model parameter values. These parameters with high uncertain values

232

Hydraulic Fracturing and Well Stimulation

Table 11.1 Estimated reservoir and fracture properties in the Naricual-5 formation. Horizontal permeability kx

0.08

md

Horizontal permeability ky

0.08

md

Fluid viscosity

8

cp

Oil formation volume factor

1.1

rb/stb

Wellbore radius

0.328

ft

Number of fractures

6

Fracture length

20

ft

Fracture height

4.5

ft

Fracture width

0.1

inch

Fracture permeability

1000

md

Reservoir pressure

12472

psia

Wellbore pressure

5339

psia

include the number of fractures, fracture permeability, and fracture width. A sensitivity analysis was carried out using the baseline data in Table 11.1. Figure 11.2 shows the effect of the number of fractures on well productivity, while all other parameter values remain unchanged. It indicates that well productivity increases non-linearly with the number of fractures. This is because the drainage distances between fractures get shorter as the number of fractures increases. The rate of increase gets higher as the number of fractures grows. Thus, the optimal design of refracturing should consider creating a complex fracture network in order to maximize well productivity. Equation (11.3) implies that fracture width and fracture permeability appears in the model as a product called fracture conductivity (wkf). Therefore, there is no need to analyze the effect of each of the parameters. Figure 11.3 illustrates the effect of fracture conductivity on well productivity while all other parameter values are kept constant. It demonstrates that well productivity increases non-linearly with fracture conductivity. The benefit of increasing fracture conductivity levels out beyond 2000 md-inch. Therefore, there is no need to create high-conductivity fractures in refracturing operations.

Productivity of Refractured Oil Wells 233

Well production rate (stb/day)

250 200 150 100 50 0 0

2

4 6 Number of fractures

8

10

Figure 11.2 Calculated effect of number of fractures on well productivity.

Well production Rate (stb/day)

350 300 250 200 150 100 50 0 0

2000

4000 6000 8000 Fracture conductivity (md-inch)

10000

Figure 11.3 Calculated effect of fracture conductivity on well productivity.

11.5 Conclusions Assuming uniformly distributed radial fractures around wellbore, a semi-analytical model was derived in this study to predict productivity of refractured oil wells under a pseudo steady state flow condition. Field case study and sensitivity analyses were performed with the semi-analytical model. The following conclusions are drawn: 1. Results of a field case study for a well completed with high energy gas fracturing indicate that the mathematical model

234

Hydraulic Fracturing and Well Stimulation over-estimates well productivity by about 10%. This inconsistency is attributed to the uncertainties in estimating model parameter values. 2. Sensitivity analysis with the mathematical model shows that the productivity of refractured wells increases non-linearly with the number of radial fractures around the wellbore. This is because the drainage distances between fractures get shorter as the number of fractures increases. Because the rate of increase gets higher as the number of fractures grows, optimal design of refracturing should consider forming a complex fracture network. 3. Sensitivity analysis with the mathematical model also shows that the productivity of refractured wells increases nonlinearly with fracture conductivity, but the benefit of increasing fracture conductivity levels out beyond 2000 md-inch. Therefore, there is no need to create high-conductivity fractures in refracturing operations.

Acknowledgements The authors are grateful to Chevron USA and the UNOCAL Corporation for providing the LA Board of Regents Chevron I and Chevron II Professorships in Petroleum Engineering and UNOCAL Professorship in Engineering throughout this study.

References 1. R. Khusainov, B. Ganiev, A. Karimova, and O. Karpova, Refracturing is the best way to develop hard-to-recover reserves in Romashkino oilfield conditions. Presented at SPE Russian Oil and Gas Exploration and Production Technical Conference and Exhibition, Moscow, Russia, 14–16 October (2014). 2. E. Urban, D. Orozco, A. Fragoso, K. Selvan, and R. Aguilera, Refracturing Vs. infill drilling - A cost effective approach to enhancing recovery in shale reservoirs. Presented at Unconventional Resources Technology Conference, San Antonio, Texas, USA, 1–3 August (2016). 3. M. C. Vincent, Refracs: Why do they work, and why do they fail in 100 published field studies? Presented at SPE Annual Technical Conference and Exhibition, Florence, Italy, 19–22 September (2010).

Productivity of Refractured Oil Wells 235 4. N. P. Roussel and M. M. Sharma, Quantifying transient effects in alteredstress refracturing of vertical wells. SPE J. 15, 770–782 (2010). 5. N. P. Roussel and M. M. Sharma, Selecting candidate wells for refracturing using production data. SPE Production & Operations 28, 36–45 (2013). 6. J. Ren, J. Guo, Y. Deng, L. Zhang, and H. Lin, Refracturing strategy for sevensprings oilfield in china based on geological and engineering characteristics. Presented at SPE/IATMI Asia Pacific Oil & Gas Conference and Exhibition, Nusa Dua, Bali, Indonesia, 20–22 October (2015). 7. G. Miller, G. Lindsay, J. Baihly, and T. Xu, Parent well refracturing: economic safety nets in an uneconomic market. Presented at SPE Low Permeability Symposium, Denver, Colorado, USA, 5–6 May (2016). 8. R.G. Argawal, R.D. Carter, and C.B. Pollock, Evaluation and prediction of performance of low-permeability gas wells stimulated by massive hydraulic fracturing. J. Petrol. Technol. 31, 362–372 (1979). 9. H. Cinco-Ley and F. Samaniego, Transient pressure analysis for fractured wells. J. Petrol. Technol. 33, 1749-1766, (1981). 10. B. Guo and D.S. Schechter, A simple and rigorous IPR equation for vertical and horizontal wells intersecting long fractures. J. Can. Petrol. Technol. 38, 46–54 (1999). 11. J. Li, B. Guo, D. Gao, and C. Ai, The effect of fracture face matrix damage on productivity of fractures with infinite and finite conductivities in shale gas reservoirs. SPE Drilling & Completion 27, 347–353 (2012). 12. J. Ramirez, J. Barrera, and R. Romero, Propellant-assisted perforating in high-pressure and temperature wells at Campo Bosque in Northern Monagas State. Presented at SPE Annual Technical Conference and Exhibition, New Orleans, Louisiana, USA, 30 September–3 October (2001).

Appendix A: Derivation of Infow Equation for Wells with Radial Fractures under Pseudo-Steady State Flow Conditions Radial fractures around a vertical wellbore are illustrated in Figure 11.1. The following assumptions are made in deriving a well inflow model: 1. The rock in the pay zone is homogeneous. 2. Reservoir pressure is above the bubble point so that single phase liquid flow dominates. 3. Linear fluid flow governed by Darcy’s Law prevails in the rock. 4. All fractures are identical in geometry. 5. Fluid flow from the rock directly to the wellbore is negligible. 6. Pseudo-steady state flow prevails in the fractured region.

236

Hydraulic Fracturing and Well Stimulation y

p y

r da

un

ow -fl

Bo

No

o

a

r

Lf

A

x

x Δx

Fractures

Figure A11.1 A plan view of fractures around a wellbore.

Figure A11.1 shows a plan view of radial fractures around a wellbore. It can be shown that the half-angle between fractures is a = π/n, where n is the number of fractures. Consider the fluid flow from point P to point A at a radial distance, x, from the wellbore centerline. The fluid flow rate, q(x), over a short interval of fracture dx can be formulated. Under a pseudo-steady state flow condition, the fluid production is driven by the expansion of fluid and rock in the reservoir. The total compressibility of the reservoir is defined as:

ct

V p

1 V

(A11.1)

Differentiation of Eq. (A11.1), with respect to time, gives an expression of the flow rate

ctV

p t

V t

q( x )

(A11.2)

Productivity of Refractured Oil Wells 237 where the fluid volume is expressed as

V

x tan( ) ( x Δx )tan( ) Δh f 2

h f tan( )xΔx (A11.3)

The pressure decline rate ∂p/∂t in Eq. (A11.2) is then expressed as

p t

q( x ) ctV

q( x ) ct h f tan( )xΔx

(A11.4)

It is well known that the following equation governs linear flow in porous media: 2

p

y

2

ct p k t

(A11.5)

where the effective permeability in the y-direction, perpendicular to the fracture orientation, can be estimated based on directional permeabilities, kx k y i.e. k Substituting Eq. (A11.4) into Eq. (A11.5) yields: 2

p

y

2

q

(x )

kh f tan( )xΔx

(A11.6)

Integrating Eq. (A11.6) one time yields:

p y

q( x ) y c1 kh f tan( )xΔx

(A11.7)

where c1 is an integration constant and can be determined using the no-flow boundary condition

p y

0. y x tan( )

(A11.8)

238

Hydraulic Fracturing and Well Stimulation

Applying Eq. (A11.8) to Eq. (A11.7) gives

c1

q( x ) kh f Δx

(A11.9)

Substituting Eq. (A11.9) into Eq. (A11.7) gives

p y

q( x ) y 1 kh f Δx x tan( )

(A11.10)

Separating variables, Eq. (A11.10) is changed to

dp

q( x ) y 1 dy kh f Δx x tan( )

(A11.11)

y2 2 x tan( )

(A11.12)

which is integrated to get

p

q( x ) y kh f Δx

c2 .

where the integration constant, c2, can be determined using the boundary condition at the fracture face

p

y 0

p f ( x ).

(A11.13)

where pf(x) is the pressure in the fracture at point x. Applying Eq. (A11.13) to Eq. (A11.12) gives

c2 = pf(x)

(A11.14)

Substituting Eq. (A11.14) into Eq. (A11.12) results in

p

p f (x )

q( x ) y kh f Δx

y2 2 x tan( )

(A11.15)

Productivity of Refractured Oil Wells 239 Along the no-flow boundary, y = x tan(a), the pressure is pe. Eq. (A11.15) demands

2kh f Δx

q( x )

x tan( )

[ pe

p f ( x )]

(A11.16)

The fluid velocity in the matrix in the y-direction at the fracture face at point x can be expressed as

v( x )

q( x ) h f Δx

2k [ pe x tan( )

p f ( x )]

(A11.17)

The cumulative flow rate of fluid collected by a fracture interval between well-bore and point A can be determined based on v(x) as Lf

x

Q( x ) 2 v( x )h f dx

2

x

Lf

2kh f x tan( )

[ pe

p f ( x )]dx

(A11.18)

If the average width of the fracture is w, Darcy Velocity, vf(x), in the fracture can be formulated by dividing Eq. (A11.18) by the cross-sectional area of the fracture:

v f (x )

Q( x ) wh f

(A11.19)

Applying Darcy’s Law to the flow along the fracture gives

v f (x )

k f dp f ( x ) dx

(A11.20)

Coupling Eq. (A11.20) and Eq. (A11.19) yields

Q( x ) wh f

k f dp f ( x ) dx

(A11.21)

240

Hydraulic Fracturing and Well Stimulation

Substituting Eq. (A11.18) into Eq. (A11.21) and rearranging the latter gives x

dp f ( x )

2

dx

Lf

2k [ pe wk f x tan( )

p f ( x )]dx

(A11.22)

Differentiation of Eq. (A11.22), with respect to x, yields

d 2 p f (x ) dx

2

4k [ pe wkx x tan( )

p f ( x )]

(A11.23)

Defning pd as the pressure drawdown at point x in the fracture

pd = pe – pf (x).

(A11.24)

4k wk f tan( )

(A11.25)

and

c

Substituting Eqs. (A11.24) and (A11.25) into Eq. (A11.23) yields

d 2 pd dx 2

c pd x

(A11.26)

The first boundary condition is given at the wellbore and expressed as

pd

x rw

pd

pe

pw

(A11.27)

The second boundary condition is expressed as

pd|x

Lf

0.

(A11.28)

Productivity of Refractured Oil Wells 241 The general solution to Eq. (A11.26) is

pd

C1 x J1 (2i c x ) C2 x N1 (2i c x

(A11.29)

where J1 is a Bessel Function of the first kind in the order of 1 and N1 is a Bessel Function of the second kind in the order of 1. The constants, C1 and C2, are determined based on the boundary conditions to be

C1

pdT12 T01T11 T12T01

(A11.30)

C2

pdT11 T02T11 T12T01

(A11.31)

T01

rw J1 (2i c rw )

(A11.32)

T02

rw N1 (2i c rw )

(A11.33)

T11

L f J1 (2i c L f

(A11.34)

T12

L f N1 (2i c L f

(A11.35)

and

where

Part 5 ENVIRONMENTAL ISSUES OF HYDRAULIC FRACTURING Introduction Population growth and the vital need for energy are among important challenges to be tackled by nations across the globe. Fossil fuels have long been relied upon to meet the energy demand despite their limited quantity and environmental problems such as greenhouse gas emission which is broadly recognized as a worldwide issue encountered when heavy hydrocarbon fuels are burnt. Moving towards a more sustainable future necessitates harnessing renewable sources of energy and combining them with non-renewable sources to create a flexible energy portfolio. However, renewables are not yet ready to go online because of their high intermittency and some availability limitations. More extensive research and investigation are warranted until reaching a point where renewables will be fully able to quench the world’s thirst for energy. In a large country such as the United States, the need and desire to lessen dependency on imported conventional energy sources has always been a point of discussion. According to the United States Energy Information Administration (USEIA) the country is trending downward in the use of imported oil, after the peak year of 2005 with showing an export number greater than import in 2015 for the first time in a long time [1]. Shifting from conventional sources to the more sustainable renewables requires substantial time and effort, primarily to ensure that the energy production from renewables is economically viable [2]. Natural gas, as a candidate for short and even mid-term energy solutions, has garnered attention of both energy companies and energy demanding countries. As a light hydrocarbon fuel, natural gas is easily accessible and is dependable which means unlike some renewable energy sources, natural gas’s performance is not a function of environmental conditions such as sunlight or wind, nor is this energy source intermittent. When compared to heavier hydrocarbon fuels, natural gas emits less harmful by-products when burnt. Hydraulic fracturing can help easing the process of translation from heavy hydrocarbon fuels to a more sustainable energy resource by producing natural gas. This technology, however, is not without deficiencies when it comes to protecting the environment. Fred Aminzadeh (ed.) Hydraulic Fracturing and Well Stimulation, (243–270) © 2019 Scrivener Publishing LLC

244

Part 

Hydraulic fracturing is an old technique which has been around for several decades. Extensive use of fracturing in recent years is attributed to advances in horizontal drilling and the capability of reaching deep target formations with lower costs. Concerns have been raised on wide-spread application of hydraulic fracturing in oil and gas and other operations despite of the benefits of this technology [3, 4]. Unique safety and environmental considerations must be taken to mitigate concerns and ensure safe operations [5]. In particular, little is known about long-term environmental consequences given the relatively short history of extensive horizontal drilling. Potential environmental impacts can be classified into three major categories: air, induced seismicity, and water and wastewater. Air pollution in general is linked to natural gas development activities, including fracturing. For example, a human health risk assessment study by McKenzie et al. [6] on a gas field in Garfield County, Colorado suggested that people living at a distance greater than half a mile from the site were in less danger than the ones living at a lesser distance. According to a more recent study by Moore et al. [7], particulate matters smaller than 2.5 μm in size (PM2.5) and Nox are found in diesel emissions during preproduction (i.e. drilling and hydraulic fracturing activities) of unconventional natural gas production. Subsurface injection operations normally generate low magnitude (smaller than 2 in Richter scale) seismic events which are termed as micro-earthquakes or microseismic [8]. In a few cases of hydrofracking jobs, seismic events have been felt by nearby residents and the operator has been forced to stop the injection. In 2011, two seismic events (2.3 and 1.5 magnitude in Richter scale) were recorded close to Blackpool, Lancashire in the U.K. -- the operator had to halt the fracturing operation in nearby Bowland Shale formation [9]. Generally, the risk of generating serious earthquakes as a result of hydraulic fracturing is low when compared to the deep-well injection process, which shows higher probabilities of observed larger seismic events [8, 10, 11]. Historically, the oldest injection-induced seismic events are those of the Rocky Mountain Arsenal waste injection site in Denver, Colorado [12]. The most recent events were generated in Youngstown, Ohio [13] and central Oklahoma [14], both in 2011. The former is also known to be the largest injection-induced event with a 5.6 magnitude [8]. Potential surface water [15, 16] and groundwater contaminations are caused by either fracturing fluid chemicals or volatile compounds from deep formations [17–20]. The returned fluid handling and treatment is another source of possible environmental complications [16, 21–24].

Part 

245

Section 5 includes three chapters: Chapter 5-1 addresses major contributing factors to the safety of hydraulic fracturing. This chapter further covers analyzing different steps of a typical fracturing job and discussing root-causes of example incidents and accidents and the public perception. Post fracturing fluid recovery is the topic of Chapter 5-2 where the fate of fracturing fluid within the reservoir has been investigated and responsible factors for low recovery of large portions of injected fluid have been identified. Chapter 5-3 wraps up Section 5 by introducing and numerically solving a potential groundwater contamination problem in the domain of hydraulic fracturing where a breach in the well results in solute movement in the adjacent formation and contamination of nearby underground water resources.

References 1. U.S. Energy Information Administration, 2018. 2. Spellman, F., Environmental Impacts of Hydraulic Fracturing, CRC Press, 2012. 3. Kargbo, D.M., Wilhelm, R.G., Campbell, D.J., Natural gas plays in the Marcellus Shale: challenges and potential opportunities. Environ. Sci. Technol., 44, 15, 5679–84, 2010. 4. USEPA, Assessment of the Potential Impacts of Hydraulic Fracturing for Oil and Gas on Drinking Water Resources, 2015. 5. Jabbari, N., Ashayeri, C., Meshkati, N., Leading Safety, Health, and Environmental Indicators in Hydraulic Fracturing. SPE Western Regional Meeting, Garden Grove, CA, 27-30 April 2015, 2015. 6. McKenzie, L.M. et al., Human health risk assessment of air emissions from development of unconventional natural gas resources. Sci. Total Environ., 424, 79–87, 2012. 7. Moore, C.W. et al., Air Impacts of Increased Natural Gas Acquisition, Processing, and Use: A Critical Review. Environ. Sci. Technol., 2014. 8. Ellsworth, W.L., Science (New York, N.Y.), 341, 6142, 1225942, 2013. 9. Pater, C. De and Baisch, S., Geomechanical study of Bowland Shale seismicity, 2011. 10. Goebel, T. et al., A probabilistic assessment of waste water injection induced seismicity in central California. American Geophysical Union Fall Meeting, San Francisco, 2014. 11. Aminzadeh, F., Tiwari, A., Walker, R., Correlation between Induced Seismic Events and Hydraulic Fracturing activities in California. American Geophysical Union Fall Meeting, San Francisco, 2014. 12. Healy, J. and Rubey, W., The Denver earthquakes. Science (New York, N.Y.), 161, 3848, 1301–1310, 1968.

246

Part 

13. Ohio Department of Natural Resources, Preliminary Report on the North Star 1 Class II Injection Well and the Seismic Events in the Youngstown, Ohio, Area, 2012. 14. Holland, A., 2011. 15. Hammer, R. and VanBriesen, J., In fracking’s wake: new rules are needed to protect our health and environment from contaminated wastewater, Natural Resources Defense Council, 2012. 16. VanBriesen, J., Wilson, J., Wang, Y., 2014. 17. Osborn, S.G. et al., Methane contamination of drinking water accompanying gas-well drilling and hydraulic fracturing. Proc. Natl. Acad. Sci. U.S.A., 108, 20, pp.8172–6, 2011. 18. Vengosh, A. et al., The Effects of Shale Gas Exploration and Hydraulic Fracturing on the Quality of Water Resources in the United States. Procedia Earth Planet. Sci., 7, pp.863–866, 2013. 19. Llewellyn, G.T. et al., Evaluating a groundwater supply contamination incident attributed to Marcellus Shale gas development. Proc. Natl. Acad. Sci., 112, 20, pp.6325–6330, 2015. 20. Jabbari, N., Aminzadeh, F., de Barros, F.P.J., Hydraulic fracturing and the environment: risk assessment for groundwater contamination from well casing failure. Stochastic Environ. Res. Risk Assess., 31, 1527–1542, 2017, doi: 10.1007/s00477-016-1280-0. 21. Gregory, K.B., Vidic, R.D., Dzombak, D.A., Water Management Challenges Associated with the Production of Shale Gas by Hydraulic Fracturing. Elements, 7, 3, pp.181–186, 2011. 22. USEPA, Study of the Potential Impacts of Hydraulic Fracturing on Drinking Water Resources, 2012. 23. Gordalla, B.C., Ewers, U., Frimmel, F.H., Hydraulic fracturing: a toxicological threat for groundwater and drinking-water? Environ. Earth Sci., 70, 8, 3875–3893, 2013. 24. Glazer, Y.R. et al., Potential for Using Energy from Flared Gas for On-Site Hydraulic Fracturing Wastewater Treatment in Texas. Environ. Sci. Technol. Lett., 1, pp.300–304, 2014. 25. Jabbari, N., University of Southern California, Los Angeles, California, 2016, Retrieved from http://digitallibrary.usc.edu/cdm/ref/collection/ p15799coll40/id/231045.

12 The Role of Human Factors Considerations and Safety Culture in the Safety of Hydraulic Fracturing (Fracking) Jamie Heinecke1, Nima Jabbari2* and Najmedin Meshkati3 1

Senior Student, Daniel J. Epstein, Department of Industrial & Systems Engineering, Viterbi School of Engineering, University of Southern California 2 Ph.D. Candidate, University of Southern California, Sonny Astani Department of Civil and Environmental Engineering, Viterbi School of Engineering, University of Southern California, Los Angeles, CA 3 Professor, Sonny Astani Department of Civil/Environmental Engineering, Daniel J. Epstein, Department of Industrial & Systems Engineering, Viterbi School of Engineering, University of Southern California

Abstract Hydraulic fracturing is a well stimulation method frequently used in oil and gas operations in order to facilitate the flow movement and increase production in tight and low permeability formations. Along with horizontal drilling, hydraulic fracturing has helped the United States in meeting its enormous need for energy. Despite of the advantages such as boosting economy and energy-independency, there are potential environmental and safety issues associated which rise many questions on the future of this technology. More specifically, recent accidents and incidents have heightened public and industry concerns. This paper addresses major contributing factors to the safety of fracturing by analyzing different steps of the operation and providing associated human factors and safety culture considerations. Primary human factor root-causes of three recent fracking accidents have been analyzed, and recommendations as how to proactively address these considerations in the entire hydraulic fracturing life-cycle process are proposed. Keywords: Hydraulic fracturing, fracking, safety, human factors, accident causation, safety culture

*Corresponding author: [email protected] Fred Aminzadeh (ed.) Hydraulic Fracturing and Well Stimulation, (247–270) © 2019 Scrivener Publishing LLC

247

248

Hydraulic Fracturing and Well Stimulation

12.1 Introduction The United States currently has access to some of the largest shale oil and gas reserves in the entire world. “Since 2007, discoveries of unconventional gas including shale gas have more than doubled the estimate of North American reserves to 3,000 trillion cubic feet, enough to meet 100 years of demand” [1]. Shale rock is a unique type of rock formation that lies over a mile below the surface and contains a significant amount of gas trapped inside tiny pockets within the rock [2]. Until recently, companies found it difficult to extract these precious resources using conventional drilling methods. However, the improvement of hydraulic fracturing (also known as fracking) in recent years, has allowed companies to take advantage of these reserves. Fracking allows gas to flow to the surface by pumping water under extreme pressure down into the well in order to fracture the rock formation. The process behind hydraulic fracturing is not entirely new, but became especially advantageous when combined with another technique called horizontal drilling. This combination is so effective that according to the U.S. Energy Information Administration (EIA), production of natural gas will meet consumption needs by 2019 and 12% of all production will be exported by 2040 [3], as shown in Figure 12.1. In fact, it is primarily due to fracking in the U.S. that already, according to a recent article in the Wall Street Journal (August 12, 2014), “since March 2008, oil production has increased 58% and natural-gas output has risen 21%, making the U.S. the world’s largest producer of both fuels, according to federal and international agency statistics”

40

History

2011

Projections

Net exports, 2040 (12%) Total production

30

Net imports, 2011 (8%) Total consumption

20 10 Net imports

0 -10 1990

2000

2010

2020

2030

2040

Figure 12.1 Total U.S. natural gas production, consumption, and net imports in the reference case, 1990–2040 (trillion cubic feet) [3].

The Role of Human Factors Considerations 249 (emphasis added); and “jobs directly related to oil and gas production have nearly doubled in the past 10 years to 697,000, government data shows” [4]. Fracking will become an increasingly important topic of discussion as it starts to play a larger role in America’s energy future. Human factors consideration should be the foundation of this rapidly growing technology to ensure that it is implemented correctly in order to minimize the drawbacks of the fracking process. The potential benefits of fracking could be enormous for the United States, but only if executed properly. Due to the increasing pressure to find alternative energy sources, hydraulic fracturing is being implemented in a greater capacity than ever before. “In a decade, shale gas has risen from 2 percent of US natural gas production to 37 percent. The US has overtaken Russia as the world’s largest natural gas producer” [5]. However, such rapid and enormous growth poses significant safety risks. As companies rush to capitalize on this opportunity, they tend to over-look many safety and environmental concerns such as increased seismicity as a result of waste injection [6] and groundwater contamination [6]. These challenges have been crucial enough in some cases to make the cities take serious actions to regulate, monitor, and even halt the fracking operations just like the cases of Pavillion, Wyoming in 2009 [7] and New York in 2010 [8]. Furthermore, according to the National Institute of Occupational Safety and Health (NIOSH) [9], “the oil and gas extraction industry has an annual occupational fatality rate of 27.5 per 100,000 workers (2003–2009) - more than seven times higher than the rate for all U.S. workers. The oil and gas extraction industry employed approximately 435,000 workers in 2010. The annual occupational fatality rate in this industry is highly variable; this variation is correlated with the level of drilling activity in the industry. Fatality rates are higher when there is an increased number of active drilling and workover rigs.” An accident in drilling-related operations, like in other contexts, can be characterized as “an error with sad consequences” [10]. Nevertheless, a correct understanding of the root-causes of the aforementioned so-called “error” in terms of instances of human-machine, human-task, and/or human-organization mismatches can be greatly contributed to its prevention [11–14]. To minimize accidents which could result in fatalities and nonfatal injuries and the possibility of other negative effects associated with fracking, one must consider the human factors issues involved in each stage of the process. By examining each stage of hydraulic fracturing, the specific human factors issues involved in the overall process can be addressed and ultimately both the safety and the efficiency of the entire system can be improved.

250

Hydraulic Fracturing and Well Stimulation

As discussed by [15], the fracturing operation is known to have several challenges to the environment and also to the society ranging from groundwater contamination and air pollution to increasing traffic and incidents associated. However, the purpose of this paper is to shed light on the root causes of accidents and to analyze potential risk areas that could lead to future accidents. This paper does not extensively examine the environmental impacts of fracking, but instead it focuses on the safety of workers and well integrity from a human factors perspective. It will also analyze the human, technical, and organizational factors that contribute to well control related issues and worker safety oversight.

12.2 Benefits of Hydraulic Fracturing According to experts, shale gas has the potential to “dramatically alter the energy supply picture for North America and potentially the world as other countries are just beginning to determine the extent of their own unconventional resources” [16]. The United States has the opportunity to outperform the rest of the world if it can improve fracking techniques by fixing often overlooked human factors related issues. As the world’s leader in this technology, the United States would generate an immense stimulus. Current operations have already shown positive economic effects on local industries; “the most immediate has been in employment – more than 1.7m jobs have been created” [5]. Expanding operations will greatly help a struggling American middle class while easing our energy dependence and reducing carbon emissions (natural gas is a light hydrocarbon with fewer negative consequences than the heavier hydrocarbons that are more commonly used for fuel). Despite such potential benefits, fracking has drawn significant criticism, largely as a result of human factor issue oversight.

12.3 Common Criticisms Fracking is not without imperfections and can be improved upon to reduce the risks associated with the process. Media discussions about fracking tend to focus on the possibility of long-term risks: Despite the arguable benefits of fracking, some public health officials, environmentalists and scientists are not convinced that it is worth the potential risks. Many groups are concerned over the lack of long-term research into the effects the chemicals used in hydraulic fracking have

The Role of Human Factors Considerations 251 on the people and environment around it, as well as some grey areas in industry regulation [16].

The majority of critiques raise concerns over the contamination of fresh water supplies and argue that, “despite the industry’s claims that hydraulic fracturing is a safe and proven technology […] there have been many allegations that hydraulic fracturing had led to the contamination of drinking water in many communities” [2]. Residents in communities located in Wyoming, New York, and Colorado have insisted that their water tastes and smells abnormal since fracking began in the local area. In response, fracking companies claim that it is impossible that contaminants were able to reach the aquifer level because the fracturing process is thousands of feet away from the water table. Through numerical simulations it has been shown that in some extreme cases and rare events, failure in the system may result in contaminating the formation nearby the well and the groundwater aquifer eventually [17]. Critics also claim, in addition to water contamination, that fracking has caused more earthquakes in the area. However, while fracking does create seismic activity, “the energy released by rock breaking during shear fracturing has a magnitude about one hundred thousand times smaller than the magnitude of the slightest ‘felt’ earthquake (magnitude ~3.0) (USGS calculator)” [18]. The cause of these earthquakes is not the act of drilling and fracturing, but rather has been traced to the disposal wells used to store the contaminated flow-back fluid (e.g. [5]). “These wells are located thousands of feet underground, encased in layers of concrete and usually store the waste from several different wells” [19]. The wells can cause faults to slip resulting in an earthquake. The Rocky Mountain Arsenal in Colorado and north-central Arkansas are two well-documented cases in areas which link fluid injection disposal wells with earthquakes [20]. Both areas have seen an increase in the size and frequency of earthquakes near the underground wastewater disposal wells. Only a handful of the thousands of disposal wells have been linked with seismic activity, but because of the magnitude of this possible risk, new protocols should be established to deal with the disposal of waste fluid. For example, many companies are beginning to send their waste to a water treatment facility. Although this is likely a more expensive option, it seems to provide a promising alternative to underground disposal wells. Microseismic monitoring and data gathering during the fracturing operation [21], applications of sophisticated reflection seismology methods [22] and more specifically using real-time microseismic monitoring accompanied by keep tracking of pore-pressure changes [15] can play crucial roles in mitigating the issue and in increasing the safety of the hydrofracking operation. Also, in a study done in

252

Hydraulic Fracturing and Well Stimulation

the San Joaquin Basin of California, it has been shown that the seismic events related to the oil and gas activities can be distinguished from the natural events using semiology measures such as fractal dimension and b-value [23]. The final major area of concern for critics is the composition of fracking fluids used to break open the shale formation. Although many critics argue these fluids include harmful chemicals, “evaluating the relative volumes of the components of a fracturing fluid reveals the relatively small volume of additives that are present. Overall the concentration of additives in most slickwater fracturing fluids is a relatively consistent 0.5% to 2% with water making up 98% to 99.2%” [24]. However, it is important to remember when analyzing these percentages that millions of gallons of water are used in each drilling operation and the long term effects of what these chemicals will do in the ground is unknown. Furthermore, the exact composition of the fracking fluid used by each company is not fully revealed because it is considered proprietary, which causes additional unease. These risks and benefits of fracking are affected by operations in each stage and can be improved through human factors management. In the end, if these allegations are true, experts say that it is not due to the scientific fracking technique but instead is a result of human error. This illustrates the key obstacle fracking must overcome: the science behind fracking is valid, but in accounting for human factors, it becomes a much less reliable process that may be prone to significant and costly problems. Also, it is worthwhile to note that the concerns and worriedness about the fracturing are important issues to be addressed through communications with the public media. It, thus, necessitates proactively conducting sessions and talks on the scientific facts regarding the hydraulic fracturing and try to stay away from the myths [25–27].

12.4 Different Steps of Hydraulic Fracturing and Proposed Human Factors Considerations Hydraulic fracturing operation includes different steps starting from drilling vertically and horizontally all the way to injection and returning the fluid from the well. On the other hand, human factor considerations are vital to any system, but they are especially important when a technique is still being perfected. The improvements made by studying the human factors of the operation could allow fracking projects to realize their enormous potential. Figure 12.2 briefly discusses major steps and

The Role of Human Factors Considerations 253 Fracking Steps Vertical Drilling

Vertical Casing

Horizontal Drilling

Horizontal Casing

Description

Examples of Human Factors Considerations1

Well is spudded vertically up to a specified depth.

Using proper lifting and material handling techniques. Human-machine interface Communications Monitoring/Supervisory Control Job hazards (identification & conrol); PPEs; (”Electricity hit”) [29]

Vertical casing is inserted in place to function as a barrier to the surrounding environment.

Standard operating procedures

By reaching the kick-off point, drilling bit starts deviating from the vertical direction, creates a bend, and advances in horizontal direction

Drilling operation and equipment monitoring Using proper lifting and handling techniques. Human-machine interface Communications Monitoring/Supervisory Control Job hazards (identification & control); PPEs; (”Electricity hit”) [29]

Horizontal casing is placed after the horizontal drilling is finished. Horizontal section in entirely located in the target formation.

Following cementing standards and protocols Information monitoring (”Supervisory Control”)

Perforation

Horizontal casing is perforated using electrical charges.

Monitoring/information processing/signal detection. Situational awareness

Injection

Fracking slurry is injected through an injection pipe inside the casing and formation is cracked.

Ensure the blow-out preventer system functions properly. Information processing/monitoring Decision making

Returning Waste

Wastewater Injection

Part of the injected fluid is returned back to the surface and is temporarily stored onsite. The waste is dumped later on.

Wastewater is injected underground through a number of deep injection wells.

Ensure waste is handled properly according to regulations. Situational awareness and continuous microseismic monitoring of formation condition and integrity to avoid triggering seismic events [5] Continuous motioning and characterization of waste health hazards Standard Operating Procedures/PPEs Total system comprehension and shared mental model of the underground

1Will be further discussed in the section of hydraulic fracturing process as well as the section of human factors

Figure 12.2 Hydraulic fracturing steps (inspired from [28]) and related major human factors.

important human factors issues pertained to each step. Safety culture and its associated elements (or traits) such as organizational leadership, communication, training, and fatigue management, are cross-cutting phenomena that potentially affect every single step of the fracturing operation.

254

Hydraulic Fracturing and Well Stimulation

12.5 Hydraulic Fracturing Process: Drilling The beginning of the fracking operation involves a drilling stage. In this stage, a drill bit mounted on the end of a drill pipe creates the well. The hole extends to a level just below the aquifer zone and a casing (termed as surface casing) is inserted into the hole. Cement is pumped between the casing and the hole in order to seal off the wellbore from the freshwater. This process is continued further down the hole until there are multiple layers of cement to prevent any leakage into the water supply. When the drill reaches the appropriate length, it is then drilled horizontally into the shale formation (as shown by Figure 12.3). Casing is placed and cement is again pumped the full length of the wellbore. During this drilling process, the drill bit requires maintenance which is called “tripping pipe.” This is a multiple step job that could be potentially extremely dangerous for workers. The first stage of tripping pipes involves setting slips that attach around the stem of the drill. Generally, there are three workers involved in this process. The potential hazards of this stage include getting fingers or other body parts pinched between the slips or slip handles in addition to a risk of muscle strain due to poor lifting techniques. It is imperative for worker safety that they are trained to use proper hand placement when handling the slips and use proper lifting techniques to prevent possible injury [31]. The next stage involves setting the kelly (large hexagonal steel structure) over the hole. During this process there are three identifiable possible

Ground Water Aquifers Conductor Casing Surface Casing Intermediate Casing Production Casing Vertical Well

Horizontal Well 500 ft. Radius

Vertical Fractures in Vertical Wells Producing Formation

Vertical Fractures in Horizontal Well

Figure 12.3 View of vertical and horizontal drilling [30].

The Role of Human Factors Considerations 255 hazards. The first is the possible release of excess drilling mud which can get on the workers skin or create a slippery floor surface. In order to counter this, it is important to follow the proper protocol of shutting down the mud pumps and using a kelly that has a mud saver valve. The second hazard is the possibility of being struck by slip handles when the drill string is spun. There are alternative technologies such as a pipe spinner that companies should consider investing in to increase worker safety. The last hazard is the possibility of getting struck by the kelly when it is being placed over the hole. Again, proper training can prevent any incidents of this hazard occurring [31]. The third stage requires workers to latch elevators onto the pipe to prepare for lifting or lowering the pipe. Potential risks include getting hands pinched, caught, or being struck by elevators. Supervisors need to make sure that workers are correctly following latching procedures and elevators should be inspected and maintained after each use. Tripping the pipe also requires the worker to climb the oil derrick and work from an area called the monkeyboard (component 4 of Figure 12.4). There are numerous hazards to address in this stage. Workers can fall when they are climbing the ladder or while working from the monkeyboard. It is very important for the worker to use a climb assist device and to wear a body harness to protect against falls. They should be wearing the proper protective clothing including a hard hat, gloves, and safety-toed footwear. Companies need to ensure that slip-resistant coatings are put on all working surfaces. Additional hazards include getting caught between the pipe and various objects and being struck by dropped objects. Workers need to exercise additional caution when work is being done over head and to use proper hand placement [31]. The next stage involves using a pair of tongs to break out and disconnect the pipe. Tongs are tools that consist of two long arms that aid in seizing or holding an object. If these tongs slip or backlash during operations, serious injury could occur to the worker. It is important to implement proper break out procedure and workers should be outside of the 4 foot tong rotation area (as seen in Figure 12.5). Communication is essential between the driller and the floor hands that are operating the tongs [31]. The final stage is to rack the pipes. Crew members need to be wary of getting their hands pinched between pipes or getting feet crushed under a stand of pipe. It is important for workers to use proper hand positioning and to properly position their feet away from pipe stands. The risks in each process described above may be mitigated through human factors consideration. In the drilling stage, proper training and executing proper lifting and handling techniques will eliminate many risks.

256

Hydraulic Fracturing and Well Stimulation

Drilling Rig Components*

1. Crown Block and Water Table 2. Catline Boom and Hoist Line 3. Drilling Line 4. Monkeyboard 5. Traveling Block 6. Top Drive 7. Mast 8. Drill Pipe 9. Doghouse 10. Blowout Preventer 11. Water Tank 12. Electric Cable Tray 13. Engine Generator Sets 14. Fuel Tanks 15. Electric Control House 16. Mud Pump 17. Bulk Mud Components Storage 18. Mud Pits 19. Reserve Pits 20. Mud Gas Separator 21. Shale Shaker 22. Choke Manifold 23. Pipe Ramp 24. Pipe Racks 25. Accumulator

Equipment used in drilling

Figure 12.4 Drilling rig components [26].

Hazardous Area Mouse Hole Hole Center

RADIUS 4 FEET

Caution Area

Figure 12.5 Drilling rig floor hazardous area layout with tong swing radius [31].

The Role of Human Factors Considerations 257 Due to the fracking boom, more and more inexperienced workers are being hired and put to work without sufficient training. This puts both the workers and the environment at risk. Companies must invest in thorough training to keep their workers safe. Such training is also in the company’s best interest because the reduction of personal hazards improves job performance.

12.6 Hydraulic Fracturing Process: Fluid Injection The fluid injection stage can begin once the drilling is complete. First, the casing along the horizontal portion of the well must be perforated. To do this, a perforating gun is sent down to the desired location and a small explosive charge is set off using an electrical current. The charge punctures the well casing and part of the shale rock creating fissures to allow the fracture fluid to enter the rock formation (Figure 12.6). Fracking fluid consists of roughly 90% water, 9.5% sand, and 0.5% other chemicals. These “other chemicals” have been a topic of controversy as well. Although the percentage is small, there are millions of gallons of fracking fluid being used so the overall impact is significant. This fluid is pumped under high pressure down into the wellbore creating fissures in the shale formation. These fissures allow the trapped gas to flow into the wellbore and up to the surface for capture [32]. Workers insert a plug into the wellbore and the process is repeated multiple times throughout the horizontal wellbore, which can be over a mile long. When this is complete, the plugs are removed and the wastewater is pulled back up to the surface and stored in open pits until it can be safely disposed of. The sand from the fracking fluid prevents the fissures in the shale formation from closing,

Perforating Gun

Cement

Detonation Cord

Jet Charge

Casing

Figure 12.6 Components of the well perforation process [30].

Formation

258

Hydraulic Fracturing and Well Stimulation

creating pathways for the gas to continually flow into the wellbore and up to the surface to be harvested. Each stage of this fracking process presents a few key human factor issues that must be addressed.

12.7 Fracking Fluid The first topic that should be addressed from the human factors perspective is the fracking fluid. Fracking fluid contains powerful chemicals including hydrochloric acid, formaldehyde, and ethylene glycol. However, the real threat to worker health is the sand. “NIOSH’s recent field studies show that workers may be exposed to dust with high levels of respirable crystalline silica during hydraulic fracturing. Respirable crystalline silica is the portion of crystalline silica that is small enough to enter the gas-exchange regions of the lungs if inhaled; this includes particles with aerodynamic diameters less than approximately 10 micrometers” [31]. NIOSH findings indicate that of 116 samples collected, 47% showed silica exposures higher than the OSHA permissible exposure limit and 79% were above the recommended exposure limit. This exposure to silica is dangerous for workers because “breathing silica can cause silicosis. Silicosis is a lung disease where lung tissue around trapped silica particles reacts, causing inflammation and scarring and reduces the lungs’ ability to take in oxygen” [33]. Because these workers are forced to breathe silica repeatedly they “are at a greater risk of developing silicosis. Silica can also cause lung cancer and has been linked to other diseases, such as tuberculosis, chronic obstructive pulmonary disease, and kidney and autoimmune disease” [34]. Some possible solutions to limit the exposure of silica are to limit the distance sand falls through the air to reduce dust kick up, limit the time spent in high risk areas, and install a dust collection system on the machines that release dust. Providing respirators will also help reduce the amount of exposure the worker’s lungs receive. Proactive measures to protect employees should be a primary concern for gas companies. Not only is it extremely costly for companies to cover missed work, medical bills and lawsuits, but they will also foster greater public support by showing a concern for their workers’ well-being, thereby improving the fracking industry’s reputation.

12.8 Wastewater The second human factors topic associated with the fracking process is the flow back wastewater. Flow back fluids contain not only the harsh

The Role of Human Factors Considerations 259 chemicals found in fracking fluids but also other contaminants and naturally occurring radioactive materials from exposure to the shale formation. The fluid is usually stored in temporary pits or steel tanks until it can be properly treated or disposed of. Correctly containing these fluids within a lined pit is especially important for reducing the risk of contaminating shallow ground water, a common concern among critics: The failure of a tank, pit liner, or the line carrying fluid (“flowline”) can result in a release of contaminated materials directly into surface water and shallow ground water. Environmental clean-up of these accidentally released materials can be a costly and time consuming process. Therefore, prevention of releases is vitally important [24].

This stage possesses the highest likelihood of contaminating fresh ground water due to human error. Luckily, new systems such a closed loop fluid handling systems are being implemented throughout the industry to avoid the use of open air pits. They work by “keeping fluids within a series of pipes and tanks throughout the entire fluid storage process. Since fluid is never in contact with the ground, the likelihood of groundwater contamination is minimized” [24]. As this stage is the most at risk, there needs to be strict standards set in place to reduce the risk of any spills due to mishandling or human errors.

12.9

Human Factors and Safety Culture Considerations

Human factors and safety culture have been implicated as critical contributing factors to major accidents in both up- and down streams in the oil and gas industry. BP Deepwater Horizon (2010) [35–37] and BP Texas City Refinery fire and explosion (2005) [38] are among the famous examples of incidents. Operations of every major step in the aforementioned “Fracking Steps” on Figure 12.2, which includes Vertical Drilling, Vertical Casing, Horizontal Drilling, Horizontal Casing, Perforation, Injection, and Returning Waste are sensitive to and can seriously be compromised by the human factors and safety culture considerations.

12.9.1

Human Factors

Human factors, or ergonomics, is a scientific field concerned with improving the productivity, health, safety, and comfort of people, as well as the effective interaction between people, the technology they are using, and the environment in which both must operate. While both human factors

260

Hydraulic Fracturing and Well Stimulation

and ergonomics are used interchangeably, the term ergonomics is used when focusing on how work affects people [39]. This concerns issues such as fatigue due to prolonged monitoring tasks, injuries due to unsafe workstation, and errors due to a confusing console layout. A straight-forward justification for the need for ergonomic considerations is that the technological systems are being controlled by humans; therefore, they should be designed with the human operator’s physical and psychological needs, capabilities and limitations in mind. Ergonomics can be divided into two related and complementary areas of concentration: 1- microergonomics and 2- macroergonomics. Micro- and macroergonomic approaches build upon each other and concentrate on the introduction, integration and utilization of technology, and its interface with the end-user population. Their overall objective is to improve the safety and efficiency of the intended technological system [40].

12.9.1.1

Microergonomics

Microergonomics, also called human engineering, addresses the relationship between human, equipment and physical environment. It is focused on the human-machine system level and is, for example, concerned with the design of individual workstations, work methods, tools, control panels and displays. Microergonomics includes studies of the human body sizes, known as anthropometrics, physical and psychological abilities and limitations, information processing, and human decision-making and error. It is noteworthy that, in the context of the control room environment: “The Human Error Probability (HEP) will be reduced by factors of 2 to 10 if the workstation (display and controls) are improved by the incorporation of standard human engineering concepts” [41].

Microergonomics aims to reduce incompatibilities between operator abilities and system requirements. The following sample represents additional areas of microergonomics consideration: Materials Handling, Handtool Design and Use, Machinery Design, Workstation Design, and Workplace Environment.

12.9.1.2

Macroergonomics

Ergonomics at the macro level, macroergonomics, is focused on the overall people-technology system level and is concerned with the impact of technological systems on organizational, managerial, and personnel (sub-) systems. Macroergonomics includes areas such as training, management, the planning process, information systems, internal review/inspection programs,

The Role of Human Factors Considerations 261 performance measurement systems, reward structure, initial employee qualifications assessments, and personnel selection criteria [42]. Additional areas on which macroergonomics focuses include: Job Analysis, Training, Communications, Policies and Procedures, and Organizational Design.

12.9.2

Safety Culture

The human factors considerations, although are of great importance to fracking, consist only of necessary conditions for achieving a safe and reliable operation; to make it sufficient, one should take into account safety culture. Safety culture can be a common mode of failure for such incidents as it plays a vital role in shaping a complex technological system resiliency, robustness and defense-in-depth: “Because of their diversity and redundancies, the defense-in-depth will be widely distributed throughout the system. As such, they are only collectively vulnerable to something that is equally widespread. The most likely candidate is safety culture. It can affect all elements in a system for good or ill.” [43]

The safe and efficient operation of fracking is a function of the interactions among its human (i.e., personnel and organizational) and engineered subsystems. More specifically, interactions of its Human, Organizational, and Technological (i.e., engineered) (HOT) subsystems, within their overall operational milieu – could determine the safety culture [44]. The connection of these three (HOT) subsystems, in the context of the total system, is represented in Figure 12.7. This simplified and symbolic demonstration depicts only one critical system’s reality – the role of each subsystem as a link in a chain – in the integrity of the whole system. It does not, however, show all the needed subsystems’ interactions and interrelationships. The chain metaphor is also helpful in understanding the effects of output or production load, produced by the system, on its individual subsystems. Any increase in the output level or the capacity utilization rate imposes strain on all subsystems. Obviously, the chain (system) could break down if any link breaks down. This may occur if either all the links (subsystems) are not equally strong and designed for handling the additional load, or if they are not adequately prepared and reinforced to carry the extra load in a sustainable fashion. Major accidents at complex, large-scale technological systems have been caused by break downs of the weakest links in this chain, which are most often the human or organizational subsystems.

262

Hydraulic Fracturing and Well Stimulation

Interactive Effect Human

Organization

Technology

Volume of Output Figure 12.7 Major subsystems of a technological system.

Finally, Safety culture can also be characterized as the “overlay” or direct result of sound HOT interactions that should have been incorporated during work system’s design and nurtured and maintained during operation stages. Furthermore, an organization’s safety culture, as a system composed of behaviors, practices, policies, and structural components, cannot flourish or succeed without interactions and harmony with its environment – the societal or national culture in which the organization which runs the technological system must operate. In other words, safety culture should be considered in the context of national culture, which could seriously affect its effectiveness [45, 46]. Creating and nurturing a positive safety culture basically means to instill thinking and attitudes in organizations and individual employees that ensure safety issues are treated as high priorities. An organization fostering a safety culture would encourage employees to cultivate a questioning attitude and a rigorous and prudent approach to all aspects of their job, and would set up necessary open communications between line workers and mid- and upper management. These safety culture characteristics are equally applicable both to the operating companies as well as to their cognizant/designated governmental regulatory safety agency. It is noteworthy that the Institute of Nuclear Power Operations (INPO) has recently developed a seminal guideline and code of practice for safety culture in nuclear power industry, Traits of a Healthy Nuclear Safety Culture

The Role of Human Factors Considerations 263 (2013), that could (and should) be adopted and utilized for this purpose by the fracking industry [47].

12.10 Examples of Recent Incidents Fracking accidents and incidents can have multiple, complex and interacting causes. However as depicted in the following Figure 12.8, such events can result into three major categories of losses and adverse consequences: People-related, Environment-related, and Product/Process-related. The collective and overall impact of these three types of losses on the system’s productivity can be attributed to their negative impact on resource utilization; which could be significant, as they simultaneously reduce, lower or erode the “Output”. In order to compensate for the inefficient utilization of productive resources which should have gone into the final product, the “Input” should be increasing. The ultimate result of this vicious cycle, in the short-term, is low overall fracking “Productivity Ratio” and negative “Safety, Health, Environmental, Economical, and Social Impacts”. And in the long-term, depending on the severity of the three types of losses, major regulatory and/or political ramifications can follow. Fracking mishaps and accidents

Prevention Strategies

Prevention Strategies

Human-Systems Integration (HSI) (Human factors considerations, Situational awareness, Physical and mental Workload analysis, etc. Safety culture Procedures Standard Operating Procedure (SOP) Emergency Operating Procedure (EOP)

Loss

Engineering: Equipment, Design, Material, Operation, Maintenance, Hard and soft safety barriers Regulations Process/ Product

People

Environment

Loss

Low/ Interruption / Water of production

Incidents Injuries Accidents

Loss Spills, Emissions Contamination Unintended changes in geological formations and stability

Output Productivity Ratio

= Input

Final Outcomes Safety, Health, Environmental, Economical, and Social Impacts

Figure 12.8 Risks/loses and impacts of fracking mishaps and accidents (adapted from [48]).

264

Hydraulic Fracturing and Well Stimulation

As discussed before and is also depicted in Figure 12.8, human factors and safety culture considerations comprise a major loss prevention strategy with an overarching effect on people, environment and product/process. Example 1: Exxon Mobil subsidiary spills wastewater into the Susquehanna River Recent reports of fracking issues reveal that many of the problems are not created by the technical procedure but are a result of human error. In 2010, the Dallas News reported on a fracking spill outside Pittsburgh where a storage tank was leaking water and chemicals into a nearby river. According to the article, the XTO Energy Fracking Company that was responsible for the operation was forced to pay a $100,000 fine and to implement operations changes that cost nearly $20 million [49]. However, the leak was avoidable and not an inherent problem of fracking procedures. Irresponsible human actions caused the leak: “a valve on the storage tank had been left open and the drainage plug removed,” additionally, “XTO had ordered its water recycling contractor to another drilling site in West Virginia” [49]. Such a simple error was extremely costly for this company, yet it was entirely avoidable. Forcing the water-recycling contractor to move to another project before the current one was secure represents a human factor error associated with upper management attempting to maximize profits while disregarding important safety and oversight concerns. Example 2: Report cites inadequate management of risks Poor decision-making by management is not the only possible cause of human error. As an incident in northern Alberta revealed, worker disregard can have disastrous implications. According to the report, “workers failed to recognize and properly assess a number of issues that led to the perforation and fracturing above the base of groundwater protection” [50]. These errors resulted in the crew directly fracking into underground water table where 42 cubic meters of propane gel entered an aquifer. However, the company’s response to this incident indicates that such errors were a result of complete disregard for procedures. The company’s Chief Operating Officer, Rob Morgan, added, “all of the personnel who were involved in this particular circumstance are no longer with the company” [50]. Unfortunately, this drastic incident has fueled criticisms of fracking, yet it was a result of human error. Accidentally fracking above the aquifer level is something that should never happen. As shown in Figure 12.9, the operator missed the pay zone by over 800 meters. Operators need to be correctly reading the equipment that monitors the depth levels. If lack of experience of the operator was the issue then the company should evaluate

The Role of Human Factors Considerations 265 ~50 meters NE GL Deep Monitoring Well Accidental perforations at 136 mMD

500 mMD

177 mm Surface Casing set to 606 mMD Base of Ground Water Protection 600 m KOP 697 mMD

Previous to incident perforated and hydraulically fractured intervals

Bridge Plug 1557 mMD

114 mm Production Casing Set to 2503 mMD

1000 mMD Planned Perforation Interval 1486–1486.6 mMD

Figure 12.9 Well cross-section schematic [50].

its training program. The company should also evaluate the monitoring equipment to ensure that it is ergonomically correct and easy to use. If equipment is not adjusted to ergonomic standards then the chance for operator error is much higher. Example 3: Mud spill in Pennsylvania state forest Another incident occurred in January 2010 when “an estimated 8,000 gallons to 12,000 gallons of mud used by Anadarko E&P Co. Inc. for drilling operations over-flowed at the well site due to human error […] [Luckily, the mud] didn’t spread far enough to contaminate any surface waters, ground water, or wetlands in the area” [51]. Companies should be investing in additional monitoring equipment to have assurance in each stage of the operation. By breaking down the operation they can evaluate where the problem areas lie and what they need to do in order to improve their process. Lack of monitoring equipment in addition to inexperienced operators can be devastating to not only to the company but to the environment as well.

12.11 Conclusion and Recommendations By understanding and strengthening each of these interrelationships, the safety of the hydraulic fracturing process can be improved upon. According to a report released in 2012 by the JRC (Joint Research Centre), human-related

266

Hydraulic Fracturing and Well Stimulation

causes were present in 14% of offshore accidents at the very minimum. It is likely that human-related causes were present in even more accidents, however, no systematic analysis to identify human/organizational failures were performed in the other events [52]. It would be reasonable to assume that onshore fracking accidents have similar numbers if not more. Fracking has enormous potential for the United States in our quest for cleaner energy. The economic benefits from this process are something that even the industry’s stoutest critics have a hard time denying. “Last year, a group from Yale estimated that shale gas production contributes over $100 billion to U.S. consumers annually. Jobs have been created, many landowners have benefited financially, and lower gas prices have provided relief for consumers in the form of lower heat and electricity bills. In comparison, the authors estimated that the cost of ground-water contamination could be $250 million per year, which is 1/400th the benefit” [53]. However, we should not settle for the status quo. By strengthening current regulations and improving human factors and safety culture, we can enhance the safety of the workers, reduce accidents, and minimize adverse environmental impact of fracking.

Acknowledgment Authors would like to express their sincere gratitude to Dr. Iraj Ershaghi, Dr. Donald Paul of University of Southern California (USC), and Dr. Wayne D. Pennington of Michigan Technological University for their insights and analyses concerning technical issues affecting the hydraulic fracturing process. This work, however, should not necessarily be construed as their representative position(s).

References 1. S.L. Sakmar, Shale gas developments in North America: An overview of the regulatory and environmental challenges facing the industry, SPE North American unconventional gas conference and exhibition, The Woodlands, Texas. pp. 1, 3, 7 (2011). 2. K. Barillaro, Process of Fracking, http://shalestuff.com/education/fracking/ fracking (2012). 3. J.J. Conti, U.S. Energy Information Administration, Office of Integrated and International Energy Analysis, Annual energy outlook 2013, http://www.eia. gov/forecasts/aeo/ (2013).

The Role of Human Factors Considerations 267 4. Democrats warming to the energy industry: Lawmakers weigh benefits of fracking boom against opposition from environmentalists; some Republicans are skeptical, Wall Street Journal. p. A4, August 12 (2014). 5. D. Yergin, US energy is changing the world again, Financial Times, http:// www.ft.com/intl/cms/s/b2202a8a-2e57-11e2-8f7a-00144feabdc0,Authorised=false.html?_i_location=http://www.ft.com/cms/s/0/b2202a8a-2e5711e2-8f7a-00144feabdc0.html?s iteedition=intl&siteedition=intl&_i_referer=. p. 2. (2012). 6. K.M. Keranen, M. Weingarten, G.A. Abers, B.A. Bekins, and S. Ge, Sharp increase in central Oklahoma seismicity since 2008 induced by massive wastewater injection. Science 1255802 (2014). 7. D. DiGiulio, R.T. Wilkin, C. Miller, G. Oberlev. Investigation of ground water contamination near Pavillion, Wyoming, EPA. (2011). 8. D. Rahm, Regulating hydraulic fracturing in shale gas plays: The case of Texas. Energy Policy 39(5), 2974–2981 (2011). 9. National Institute for Occupational Safety and Health (NIOSH). Oil and Gas Extraction, Inputs: Occupational Safety and Health Risks, http://www.cdc. gov/niosh/programs/ oilgas/risks.html 10. J. Reason, Human Error, p. 162, Cambridge University Press, New York, NY. (1992). 11. J. Rasmussen, The definition of human error and a taxonomy for technical system design, in New Technology and Human Error, J. Rasmussen, K. Duncan, and J. Leplat (Eds.), John Wiley & Sons, New York, NY. (1987). 12. J. Rasmussen, Human error and the problem of causality in analysis of accidents. Invited paper for Royal Society meeting on Human Factors in High Risk Situations, 28-29 June, 1989, London, England. (1989). 13. N. Meshkati. Human Factors in Large-Scale Technological Systems’ Accidents: Three Mile Island, Bhopal, Chernobyl. Ind. Crisis Q 5, 133–154 (1991). 14. N. Meshkati. Where do accidents come from? The critical role of human and organizational factors in the safety and reliability of technological systems, Rivista Tecnica dell’ANPAC (Italian Pilots’ Union Technical Publication), 2, 14–16. (2000). 15. M.D. Zoback and A.J. Arent, Shale gas: Development opportunities and challenges, the Bridge, Nat. Acad. Eng. 44(1), (2014). 16. S. Tuthill, Understanding fracking: Arguments for and against natural gas extraction, http://www.accuweather.com/en/weather-news/fracking-environ mental-health/17444221. p. 1, (2013). 17. N. Jabbari, F. Aminzadeh, F. de Barros, B. Jafarpour, Hydraulic fracturing and potential groundwater contamination risk. Presented at the 248th American Chemical Society National Meeting & Exposition, Paper Number: 315, San Francisco, CA, 10–14 (2014). 18. K.E. George, Hydraulic fracturing 101, J. Petrol. Technol. April 34–42, (2012).

268

Hydraulic Fracturing and Well Stimulation

19. T. Henry, How oil and gas disposal wells can cause earthquakes, https:// stateimpact. npr.org/texas/tag/earthquake/.p. 1, (2012). 20. S. Horton, Disposal of hydrofracking waste fluid by injection into subsurface aquifers triggers earthquake swarm in central Arkansas with potential for damaging earthquake. Seismol. Res. Lett. 83(2), 250–260 (2012). 21. N.R. Warpinski, Understanding Hydraulic Fracture Growth, Effectiveness, and Safety through Microseismic Monitoring, ISRM International Conference for Effective and Sustainable Hydraulic Fracturing. International Society for Rock Mechanics, (2013). 22. W.D. Pennington, Reservoir geophysics. Geophysics 66(1), 25–30 (2001). 23. F. Aminzadeh, T.H.W. Goebel, Identifying induced seismicity in active tectonic regions: a case study of the San Joaquin Basin, California. AGU Fall Meeting, Abstract: S31F-06, San Francisco, California. (2013). 24. Frac Focus, Chemical Disclosure Registry, http://fracfocus.org/ (2013). 25. F. Aminzadeh, To Frack or not to Frack. That is the question. Los Angeles Science Center Series on “Science Matters”, June 8 (2013). 26. N. Jabbari, F. Aminzadeh, F. de Barros, Hydraulic Fracturing, Points and Counter Points. Urban Water Institute 20th Annual Water Policy Conference San Diego, CA, August 16 (2013). 27. F. Aminzadeh and D. Paul, Induced Seismicity Intersection of Science, Public and Regulation, IOGCC Annual Meeting, November 5. Long Beach California. (2013). 28. I. Ershaghi and Q. Qianru, Aspects of Oilfield Related Accidents. SPE Annual Technical Conference and Exhibition. Society of Petroleum Engineers. (2013). 29. M. Economides and T. Martin, Modern Fracturing: Enhancing Natural Gas Production, Energy Tribune Publishing Inc., Houston. (2007). 30. American Petroleum Institute, Hydraulic fracturing operations- well construction and integrity guidelines, API Guidance Document HF1, doi: Product No. GHF101, (2009). 31. OSHA, Oil and gas well drilling and service etool, http://www.osha.gov/ SLTC/etools/oilandgas/index.html (2014). 32. Clean Water Action, Fracking: The process, http://www.cleanwateraction. org/page/fracking-process (2013). 33. J.A. Merchant, U.S. Department of Health and Human Services, National Institute for Occupational Safety and Health, Occupational respiratory diseases, Publication No. 86-102, p. 266, (1986). 34. F.L. Rice, U.S. Department of Health and Human Services, National Institute for Occupational Safety and Health, Health effects of occupational exposure to respirable crystalline silica, Publication No. 2002-129.p. 5, (2002). 35. National Academy of Engineering and National Research Council. Interim Report on Causes of the Deepwater Horizon Oil Rig Blowout and Ways to Prevent Such Events. The National Academies Press, Washington, DC. (2010).

The Role of Human Factors Considerations 269 36. I. Ershaghi and D. Luna, Validation process for human factor in complex upstream oil and gas operations. SPE 147654, SPE Annual Technical Conference and Exhibition, Denver, CO, (2011). 37. M. Tabibzadeh and N. Meshkati, Learning from the BP Deepwater Horizon accident: Risk analysis of human and organizational factors in negative pressure test. Environ. Syst. Decis. 34(2), 194–207 (2014). 38. United States Chemical Safety and Hazard Investigation Board. Investigation Report, Refinery Explosion and Fire, BP Texas City, Texas. Report No. 200504-I-TX. (2007). 39. Eastman Kodak Company. Ergonomic Design for People at Work (Vol. 1), Lifetime Learning Publications, Belmont, California. (1983). 40. N. Meshkati, Critical Human and Organizational Factors Considerations in Design and Operation of Petrochemical Plants. Proceedings of the First International Conference on Health, Safety & Environment in Oil and Gas Exploration and Production, Society of Petroleum Engineers (SPE), 11-14 November, The Hague, The Netherlands, Volume I, 627–634 (Paper # SPE 23275). (1991). 41. A.D. Swain and H.E. Guttmann, Handbook of Human Reliability Analysis with Emphasis on Nuclear Power Plant Applications. Final Report (NUREG/ CR-1278). Washington, D.C. U.S. Nuclear Regulatory Commission. pp. 11–15, (1983). 42. H.W. Hendrick, Macroergonomics: A concept whose time has come. Human Factors Soc. Bull. 30(2), 1–3, (1987). 43. J.T. Reason, A Life in Error: From Little Slips to Big Disasters. Ashgate, Burlington, VT, (2013). 44. N. Meshkati, Human factors in process plants and facility design, in CostEffective Risk Assessment for Process Design, R. Deshotels and R. Zimmerman (Eds.), pp.113, 115, McGraw-Hill, Inc., New York, NY. (1995). 45. N. Meshkati, Cultural context of nuclear safety culture: A conceptual model and field study, in Nuclear Safety: A Human Factors Perspective, J. Misumi, B. Wilpert, R. Miller (Eds.), pp. 61–75, Taylor and Francis, London. (1999). 46. M.J. Gelfand, M. Frese, E. Salmon, Cultural influence on errors: Preventions, Detections, and Management, in Errors in Organizations, D.A. Hofmann and M. Frese (Eds.), pp. 273–315, Taylor & Francis, Routledge, New York, NY. (2011). 47. Institute of Nuclear Power Operations. Traits of a Healthy Nuclear Safety Culture. Institute of Nuclear Power Operations. INPO 12-012. http:// pbadupws.nrc.gov/docs/ ML1303/ML13031A707.pdf. (2013). 48. N. Meshkati, Preventing Accidents and Explosions in the Petroleum Industry: The Critical Role of Human Factors. Invited plenary presentation at the First International Congress on Medicine in the Petroleum Industry, organized by Pemex, the State-owned oil company of Mexico. (1992).

270

Hydraulic Fracturing and Well Stimulation

49. J. Osborne, Exxon mobil subsidiary faces criminal charges over 2010 fracking spill, Dallasnews, http://www.dallasnews.com/business/energy/20130911-exxon-mobil-subsidiary-faces-criminal-charges-over-2010-fracking-spill.ece. p. 1, (2013). 50. C. Linnitt, Alberta finds mismanagement of errors causes fracking water contamination. http://www.desmogblog.com/2012/12/22/alberta-findsmismanagement-errors-causes-fracking-water-contamination-alberta. p.1.(2012). 51. R. Swift, Spill in state forest moves gas drilling moratorium debate, TimesTribune, http://thetimes-tribune.com/news/spill-in-state-forest-moves-gasdrilling-morato-rium-debate-1.705590, p. 1, (2010). 52. M. Christou and M. Konstantinidou, Safety of offshore oil and gas operations: Lessons from past accident analysis, JRC Scientific and Policy Reports, http:// publications.jrc.ec.europa.eu/repository/bitstream/111111111/27463/1/offshore-accident-analysis-draft-final-report-dec-2012-rev6-online.pdf (2012). 53. R. Rapier, Both sides mislead when it comes to fracking, http://blogs.wsj. com/experts/2013/11/14/both-sides-mislead-when-it-comes-to-fracking p. 1, (2013).d

13 Flowback of Fracturing Fluids with Upgraded Visualization of Hydraulic Fractures and Its Implications on Overall Well Performance Khush Desai* and Fred Aminzadeh† The Mork Family Department of Chemical Engineering & Materials Science, Viterbi School of Engineering, University of Southern California, Los Angeles, California, USA

Abstract The increasing popularity of hydraulic fracturing follows some achievements in the USA, where it has become a proven technology in the stimulation of tight reservoirs. Nevertheless, the physics behind the process are not completely understood, particularly in the domain of post fracturing fluid recovery. In many instances, the recovery of large portions of injected fracturing fluid has not been successful. In this research, the goal was to identify and evaluate the responsible factors. The scope of our study includes determining the fate of the fracturing fluid within the reservoir, and calculating the loss in incremental production as a consequence of that outcome. The information can be used to more effectively predict the performance of stimulated wells with hydraulic fractures over time. Also, an estimation of incremental oil recovery post treatment can be calculated more accurately. This new knowledge is beneficial to many of the participants - service provider companies will enjoy a clear understanding of the treatment; operating companies should perform more reliable economic analyses; the individual states could realize increased accuracy in the quantification of reserves; and regulatory agencies can better determine the probability of these fluids in groundwater contamination. Keywords: Flowback, fracturing fluids, hydraulic fracturing, streamtubes, well performance *Corresponding author: [email protected] † Corresponding author: [email protected] Fred Aminzadeh (ed.) Hydraulic Fracturing and Well Stimulation, (271–283) © 2019 Scrivener Publishing LLC

271

272

Hydraulic Fracturing and Well Stimulation

13.1 Introduction A streamtube model has been developed as a tool to predict well performance post hydraulic fracturing treatment. It is also useful in ascertaining various possibilities of representative streamtube models that would be characteristic of the reservoir. As stated earlier, major portions of injected fracturing fluid cannot be recovered. Various physical phenomena impact that reality. Several sensitivity analyses have been conducted in order to correctly prioritize the contributing issues. The loss of potential production has been calculated using different input parameters resulting from damage caused by not recovering injected fracturing fluids. Exponential decline has been applied for the calculation of well production rates over time. Results were obtained for administering the simulation for a finite time period. An ongoing integrated work considering the causes of partial flowback is the in situ stress analysis in a reservoir. The purpose is to incorporate the stress distribution along the horizontal section. More importantly, the fracture closing might be a function of proppant travelling and stress regimes surrounding the wellbore. Each fracture stage will be classifed based upon stress regime at a distance from the wellbore using seismic data, and with image and sonic logs as secondary data.

13.2 Assumptions Our assumptions from this research are as follows: • • • • • • • • • •

Vertical well with PKN Fracture Geometry Formation of a single fracture Homogeneous formation and Isotropic formation No heat transfer effects Water wet rock with initial water saturation equal to irreducible water saturation Comparable properties of pad and fracturing fluid Exponential decline for producing wells Effective injection rates Negligible effects of gravity Uniform proppant distribution

Flowback of Fracturing Fluids

273

13.3 Upgraded Visualization of Hydraulic Fracturing 13.3.1

Concept

Production from any given well is the result of formation fluids being transported as a consequence of pressure differential. The path that the fluid takes within the reservoir can be visualized as a network of very small diameter tubes. Formation fluids flow within the tubes, which are known as streamtubes or flowtubes. A sizeable number of these streamtubes exist within the reservoir. It is important to note that only those streamtubes that are hydraulically connected to the wellbore contribute to well production. The flowrate of well post hydraulic fracturing treatment is dependent upon the distribution of the different sizes of streamtubes and the velocity of fluid flowing within them. Figure 13.1 shows a visualization of streamtubes. It is a Computational Fluid Dynamics (CFD) Simulation of the flow of water within a 7 mm core of Belgian sandstone. It was carried out by inCT, Materialize and TotalSim. The warmer colours in the picture represent higher fluid velocity. The upgraded visualization of hydraulic fracturing involves looking at the treatment as the rearrangement of grains. During a hydraulic fracturing procedure, the grains are rearranged so that more of the larger diameter streamtubes are created. In so doing, the area of contact between fluid and rock per unit volume of formation fluid is reduced. The result is a narrowing of boundary effects; thus, an increase in the average velocity of fluids within the tubes is observed. That increase in fluid velocity leads to an increase in the throughput, which has been realized as an enhanced production rate post hydraulic fracturing operation.

Low velocity streamtubes

7 mm

High velocity streamtubes 7 mm

Figure 13.1 CFD simulation of water within 7 mm Belgian Sandstone [7].

Hydraulic Fracturing and Well Stimulation

274

Figure 13.2 Narrowing down of capillary tubes, HiWAY by Schlumberger [8].

This concept has been applied by Schlumberger in the technology named ‘HiWAY’ for hydraulic fracturing treatments. In this technique, they attempt to distribute proppants within the fracture with specialized blending equipment that injects the proppant in pulses so that stable channels of larger diameters are created, which result in infinite fracture conductivity. Figure 13.2 shows a schematic of the concept.

13.3.2

Results

All the points in Plot 13.1 are the possibilities of the initial production rate of the well post hydraulic fracturing treatment. The magnitudes of the increase in the well production rate comply with the fact that the production rate of the well will be high in case of higher cumulative volume of connected tubes (higher reservoir permeability) and vice versa for a fixed number of set (The term ‘set’ or ‘sets’ here and onwards in this paper means the number of possibilities of range of radius of the streamtube in the distribution). Also, for a fixed volume of connected tubes the production rate 450

No increments or decrements 10% incremental unconnected tubes

Production (Bbls/day)

400

20% incremental unconnected tubes 30% incremental unconnected tubes

350

40% incremental unconnected tubes 10% decrement in connected tubes

300

20% decrement in connected tubes 250

30% decrement in connected tubes 40% decrement in connected tubes

200 150 100 6

8

10

12

14

Sets

Plot 13.1 Comparison of change in throughput.

16

Flowback of Fracturing Fluids

275

of the well would decrease with increasing number of sets. These observations validate the model. Below are observations from Plot 13.1: • Production rises by increasing the communication of natural fractures that result from the hydraulic fractures. • Production rates increase as the number of sets decrease, which is logical because a lesser number of sets would correspond to better hydraulic fracturing treatment. • Production rates increase as cumulative cross sectional area is enlarged, which is appropriate as a higher cross sectional area means more stream-tubes being hydraulically connected, and hence higher production. • The magnitude of increase or decrease in the flowrate is directly and linearly related to magnitude of the increase or decrease in the cumulative cross sectional area of hydraulically connected streamtubes. • The production rate changes non-linearly with a higher number of sets. Using the data of distribution of the streamtubes up to the drainage radius of a well, a model can be built to characterize the network of streamtubes. When using the output of that model as input for this model, the performance of that well following treatment can be predicted with reasonable certainty. In addition, this model can be used to find the representative distribution of streamtubes of the entire reservoir after the hydraulic fracturing treatment. It is challenging, as there can be varying scenarios of distributions that may show the same flowrate of the well. The possibilities can be converged by using information from other sources including, but not limited to, seismic, core analysis and logs.

13.4 Reasons for Partial Flowback 13.4.1

Fracture Modelling

The Geertsma-De Klerk (KGD) and Perkins-Kern-Nordgen (PKN) models are widely accepted two dimensional models for fracture geometry. An assumption was made wherein the length of the fracture is considerably more than the fracture height. In such a scenario, the PKN [1] model is more applicable.

276

Hydraulic Fracturing and Well Stimulation

The following equations have been used to calculate fracture length, width, and the average width with time during the treatment:

Q 3t 4 E L 51.46 (1 v 2 ) h 4

1/5

(1 v 2 )Q L Wmax 0.389 E Wavg 0.628 Wmax

1/ 4

Where, L is fracture half-length in ft. Q is injection rate in BPM t is time in minutes h is formation thickness in ft. E is young’s modulus in PSI v is poisson’s ratio which is unitless μ is viscosity in cp

13.4.2

Depth of Penetration

It is essential to know the distribution of the fracturing fluid within the reservoir to properly characterize which part of that fluid flows back. Those sections of the hydraulic fracture that are close to the wellbore will experience exposure to higher pressure for longer durations of time during the treatment. For a fracture to propagate, signifcant treatment pressures must be applied from the surface to provide sufficient pressure differential at the tip of the fracture. Farther from the wellbore, pressure is diminished due to turbulence and friction pressure. The magnitudes of pressure experienced in each section of wellbore will progressively increase as the fracture penetrates deeper in the reservoir [2]. Taking the effect of both of these aspects into consideration, a plot has been developed to obtain a visualization of distribution of fracturing fluid within the reservoir. Since the viscosity of fracturing fluid is much higher than the reservoir fluids, the relative permeability of the fracturing fluid is high. As a result, fingering takes place which leads to deeper penetration of the fracturing fluids into the reservoir. Plot 13.2 was obtained for the depth of penetration of fluids over the fracture half length.

Flowback of Fracturing Fluids

277

10 Depth of penetration (ft)

9 8 7 6 5 4 3 2 1 0 0

100

200

300

400 500 Fracture half length (ft)

600

700

800

Plot 13.2 Fracture half-length vs. depth of penetration.

13.4.3

Closing of Fractures

At the conclusion of a hydraulic fracturing treatment, the surface pumps are shut down, which causes the pressure in the fracture to decline instantaneously. Due to the existence of stresses within the rock, the fractures close off. Ideally, the segment of fracture that closes is directly dependent upon proppant distribution within the fracture [3]. For simplicity, in this model the assumption is to provide uniform distribution of proppants. The length of fracture that closes is the section where proppants have not been placed [4]. A criteria was used to find the region of fracture where proppants were unable to reach. The criteria is that proppants can penetrate in the fracture only until the width of fracture is twice the diameter of proppant. In Plot 13.2, a red line has been drawn at a distance of 133.56 feet from the end of the fracture half length. It means the amount of fracturing fluid lost on the right side of the red line is potentially unrecoverable. Its volume has been calculated to be 19.31 bbls, which is 4.29% of the volume of injected fracturing fluid.

13.4.4

Chemical Interaction of Fracturing Fluids

Once the well is producing, subsequent to hydraulic fracturing treatment, a comingled flow might occur from reservoir and fracturing fluids. After a certain amount of time, only the flow of formation fluids is contributing and dominating. From observation, it is unlikely to be due to all the fracturing fluids being produced. As production continues, the pressure

278

Hydraulic Fracturing and Well Stimulation

transience gets deeper into the reservoir. The pressure differential should be enough for the fracturing fluid to flow within the reservoir. However, fracturing fluids react with reservoir rock and fluids over time, which decrease their mobility, and thus require higher differential pressure to overcome the capillary pressure and flow within the reservoir. In reality, applying such high differential pressures in the reservoir may not be practical, leading to unproduced injected fracturing fluids within the reservoir. This phenomenon is a major factor in high volumes of fracturing fluid remaining in the reservoir. Calculated Values: • Additional irrecoverable volume of injected fracturing fluid = 455.49 bbls • Recoverable volume of injected fracturing fluids = 786.67 bbls • Total recoverable oil = 3121.86 bbls (Virgin zone) + 3540.02 bbls (Affected zone) = 6661.88 bbls

13.5 Impact of Parameters under Control The data analysis shows that approximately 45% to 80% of injected fracturing fluids are recoverable under certain conditions. From 20% to 55% of that fluid stays in the reservoir, of which 4.29% is due to the closing of fractures, and 15.71% to 50.71% is caused by the interaction of fracturing fluid with reservoir fluids and rock. It is very evident that we get a higher percentage of loss of fracturing fluid due to the interaction of those injected fluids with reservoir rock and fluids. The fracturing fluid can be more easily recovered by both decreasing viscosity of fracturing fluid post treatment and increasing the time for comingled flow. From sensitivity analysis, 344.94 bbls of incremental oil per 1 cp of decrease in fracturing fluid viscosity post treatment is achieved as per Plot 13.3. However, 5.26 bbls of incremental oil per 1 hr of increase in time of comingled flow is achieved as per Plot 13.4. This means the performance of stimulated well is highly sensitive to the viscosity of fracturing fluids post treatment. Breaker is one of the additives of fracturing fluid for decreasing its viscosity post treatment. Hence, the performance of the stimulated well is dependent on the performance of breaker.

Flowback of Fracturing Fluids

279

Additional irrecoverable volume (bbls)

5000 4500 4000 3500 3000 2500 2000 1500 1000 500 0

0

10

5

15 (cp)

20

25

30

Additional irrecoverable volume (bbls)

Plot 13.3 μ vs. recoverable oil volume.

10000 9000 8000 7000 6000 5000 4000 3000 2000 1000 0 0

200

400

600

800

1000

1200

1400

t (hrs)

Plot 13.4 Time vs. additional irrecoverable fracturing fluid volume.

13.6 Loss in Incremental Oil Production The fracturing fluid that stays back in the reservoir acts as a restriction to the reservoir fluids flowing into the fracture. As a result, there is skin associated with fluids not flowing back. That skin can cause a decrease in production rate over time and therefore lessen the cumulative oil production.

280

Hydraulic Fracturing and Well Stimulation 900 800

qo (bbls/day)

700 600 500 400 300 200 100 0 0

100

200

300

400

500

600

700

800

900

1000

Time, post comingled flow (days) Without skin due to fracturing fluid

With skin due to fracturing fluid

Plot 13.5 Decline curve.

In the case of Plot 13.5, an incremental production of 36,676 bbls of oil could be achieved over the duration of 900 days if the formation damage due to unproduced fracturing fluid had been avoided [5].

13.7 Conclusions We conclude that by combining the streamtube model approach with the knowledge of geomechanics of the field, the well performance after hydraulic treatment can be predicted with reasonable accuracy. Partial flowback of fracturing fluids is primarily a result of: • High pressure injection of viscous fracturing fluid leading to larger depths of penetration where it becomes very difficult to get enough pressure differential to overcome the capillary pressure of fracturing fluids. Furthermore, the interaction of fracturing fluid with reservoir rock and fluids substantially reduces the mobility of fracturing fluid. The situation is aggravated by high viscosity of fracturing fluids post treatment. This is a major contributor in the loss of fracturing fluid in the reservoir and its impact on overall well performance. • The closing of fractures following treatment causes the injected fluid to be placed in sections of the reservoir which are relatively hard to access. This is a minor contributor in the loss of fracturing fluid in the reservoir and its impact on overall well performance.

Flowback of Fracturing Fluids

281

In this model, 45% to 80% of injected fluid can potentially be recovered. The skin caused by unproduced fracturing fluids can prevent us from producing tens of thousands of barrels of oil over the life of the well. Wells with lower skin values are much more sensitive to the skin due to unproduced injected fracturing fluids.

13.8 Limitations All the estimates of the magnitude of loss of fracturing fluid are low when considering only one fracture. However, in reality, the estimates would increase substantially due to the complex geometry of fracture where a single fracture does not exist. In actuality, the distribution of the proppants within the fracture is not uniform. For that reason, the fracture may close off in the early or middle region, making the rest of the length of the fracture highly ineffective. In that case the loss of fracturing fluid due to closing of fractures may be much more than what has been previously calculated. Usually, the concentration of proppants is greater in the lower parts of a fracture due to the effect of gravity. One would encounter losses attributable to heterogeneity, natural fractures and faults in real life which have not been taken into account in this model. That could be one of the contributing factors for further losses of injected fracturing fluid impacting the observation of 2–26% flowback [6].

References 1. T.K. Perkins and L.R. Kern, SPE-89-PA widths of hydraulic fractures. J Pet Technol 13(9), 937–949 (1961). 2. T. Martin and P. Valko, Hydraulic Fracture Design for Pressure Enhancement, Modern Fracturing Enhancing Natural Gas Production, Houston, Texas. 3. R.A. Woodroof, M. Asadi, and M.N. Warren, SPE-82221-MS Monitoring fracturing fluid flowback and optimizing fracturing fluid cleanup using chemical Frac Tracer, SPE European Formation Damage Conference, 13–14 May, The Hague, Netherlands (2003). 4. http://www.corelab.com/ps/hydraulic-fracture-design 5. http://www.fekete.com/SAN/WebHelp/FeketeHarmony/Harmony_ WebHelp/Content/HTML_Files/Reference_Material/Analysis_Method_ Theory/Blasingame_Theory.htm 6. Q. Zhou, R. Dilmore, A.N. Kleit, and J. Yilin, SPE-173364-MS Evaluating Fracturing Fluid Flowback in Marcellus Using Data Mining Technologies.

282

Hydraulic Fracturing and Well Stimulation

7. http://biomedical.materialise.com/cases/cfd-simulation-water-flowsandstone-using-micro-ct 8. http://www.slb.com/services/completions/stimulation/sandstone/hiway_ channel_fracturing.aspx?t=2

Appendix - A Calculations for Upgraded Visualization:

Vavg

R2 p 8 x

Where, Vavg is average velocity in m/s R is radius in m μ is viscosity in kg/m/s P is change of pressure in kg/m2 x is change in distance in m Calculations for Damage due to Fracturing Fluids:

S frac

S

k r 1 ln s ks rw

Sfrac is skin due to unproduced fracturing fluid and its unitless S is total skin due to everything but unproduced fracturing fluid and its unitless k is virgin reservoirs permeability in mD ks is damaged reservoir permeability in mD rs is radius of damaged region in ft rw is wellbore radius in ft

Flowback of Fracturing Fluids Calculations for Flowrate of Well:

( Pwf

Pi )

k0

h

0

q0 162.6 BO

log

1688 Ct rw2 t

(0.868 S )

Where, q is flowrate in bbl/day Pwf is wellbore flowing pressure in psi Pi is the initial reservoir pressure in psi k0 is relative oil permeability and its unitless μ0 is oil viscosity in cp h is thickness of formation in ft Bo is the formation volume factor in rb/stb ϕ is porosity and it’s unitless Ct is the total compressibility of the reservoir in psi–1 rw is wellbore radius in ft t is time in hrs S is skin and its unitless Calculations for Change in Flowrate of Well with Time:

qt = qi * exp(–ta) Where, qi is initial production rate in stb/day qt is production rate at time t in stb/day t is time in days a is decline factor and its unitless

283

14 Assessing the Groundwater Contamination Potential from a Well in a Hydraulic Fracturing Operation Nima Jabbari1*, Fred Aminzadeh2 and Felipe P. J. de Barros1 1

University of Southern California, Sonny Astani Department of Civil and Environmental Engineering, Los Angeles, CA, USA 2 University of Southern California, Mork Family Depart of Chemical Engineering and Material Science, Los Angeles, CA, USA

Abstract The introduction of hydraulic fracturing as a new technology for shale gas production from low-permeability geologic formations has revitalized natural gas as an abundant, economic, and cleaner source of energy. However, hydraulic fracturing has raised several environmental concerns, from air pollution and drinking water contamination to on-site accidents and risks associated with potential seismic activities induced by subsurface fluid injection. Some of the potential incidents and accidents could be mitigated by adding more precautions to the different stages of an operation. Integrity of the injection well is one of the most important factors to be considered when dealing with water resources contamination. This work focuses on groundwater contamination potential in a leakage accident from the casing during hydraulic fracturing. A numerical flow and transport modeling approach is adopted to explore and quantify the associated risk of groundwater contamination under various hypothetical scenarios. The goal is to investigate a few cases with different geological and operational parameters and, also, to evaluate the importance of the well integrity, especially along the interval in which well casing is inside an aquifer. The results of the work show that if the well integrity is compromised, the groundwater will be contaminated by the chemicals in a relatively short duration. It is also shown that the result of the stochastic work is dependent on how well one knows the failure scenario. The appropriate choice of

*Corresponding author: [email protected] Fred Aminzadeh (ed.) Hydraulic Fracturing and Well Stimulation, (285–301) © 2019 Scrivener Publishing LLC

285

286

Hydraulic Fracturing and Well Stimulation

the probability distribution function for the leakage point is of high importance in simulations such as these. Keywords: Hydraulic fracturing, well integrity, groundwater, environmental

14.1 Introduction Hydraulic Fracturing is an old reservoir stimulation method that has been acknowledged in petroleum industry since fifty years ago. The technology aims at increasing the oil and gas production through a sequence of processes including drilling vertical or/and horizontal wells, injecting highly pressurized fluid, cracking the matrix formation, and creating fissures, which, in turn, increases permeability values of the reservoir [1]. As stated by Daneshy [2], hydraulic fracturing boosts the production through changing the flow pattern regime toward the well from radial to linear. At the same time, horizontal wellbores help in obtaining a larger reservoir-well contact which, in turn, results in higher production numbers. A combination of advanced horizontal drilling technologies and hydraulic fracturing has become very popular lately in producing oil and natural gas from shale and other tight formations [3, 4]. Despite the popularity and the common use, this technique still has imperfections that raise questions on the safety and environmental aspects of the process [5, 6]. In addition to the safety, environmental problems are also important since the long-term effects of the hydraulic fracturing activities are not very well understood yet [7]. This fact creates lots of questions for the environmentalist and the public. The environmental impacts can be viewed in three categories: air, water, and induced seismicity. Air pollution, in general, is linked to several different activities carried out during natural gas development including drilling and fracturing [8–10]. Subsurface injection operations normally generate low magnitude (smaller than 2 in Richter scale) seismic events which are termed as micro-earthquakes or microseismic [11]. In a few cases of hydrofracking jobs, seismic events have been felt by the nearby residents and make the operator stop the injection. As an example, in 2011 two seismic events (2.3 and 1.5 magnitude in Richter scale) were recorded near Blackpool, Lancashire in the United Kingdom which made the operator halt its fracking operation in the nearby Bowland Shale formation [12]. Generally, the risk of generation of serious earthquakes as a result of hydraulic fracturing is low when compared to the deep-well injection process, which

Groundwater Contamination Potential

287

shows higher probabilities of observing larger seismic events [11, 13, 14]. Historically, the oldest injection-induced seismic events are those of the Rocky Mountain Arsenal waste injection site in Denver, Colorado [15]. The most recent events were generated in Youngstown, Ohio [16] and central Oklahoma [11], both in 2011. The former is also known to be the largest injection-induced event with a 5.6 magnitude [11]. Chemicals used in a typical fracturing job vary and range from benign (e.g. guar gum), to toxic (e.g. Tetramethylammonium chloride), to very toxic materials (e.g. Kathon which is a biocide) [17]. Injecting these chemicals under the ground, if not practiced under safe and sound conditions, can an increase in the risk of having ground or surface water bodies contaminated. The importance of a solid well system with full integrity is pronounced here. Handling and management of returned fluid is another aspect of the fracturing operation which can potentially be a threat to the health of drinking water resources [18–21]. The well integrity issue is a quite well-known phenomenon in injection operations such as carbon sequestration and deep waste injection [22–27]. If the well integrity is not maintained, groundwater can be a target for the contaminants originating from the initial injectant or found in the returned fluid [28, 29]. From the human health point of view, groundwater pollution is critical, more specifically for regions with water shortage and high demand for groundwater tables [3, 9]. Underground failure and upward fluid migration has been listed as one of the important risk pathways to the shallow groundwater [30]. Damages in well integrity and leakage of fracking fluid into adjacent formation, poor design of the fracturing job and fracture growth in overburdened formations, reactivation of faults as permeable pathways for the fluid, and fluid flow through abandoned wells are along risk pathways discussed by the United States Environmental Protection Agency (USEPA) [3, 18]. However, among the discussed risk pathways, the chance of upward fluid migration from deep formations to the shallow aquifers is reduced as a result of a number of factors. Absence of a hydraulic connectivity between the aquifer and shale formation because of geometrical features and physical barriers, makes it unlikely for reservoir fluid to migrate upward, mostly, as stated by several authors. The separation between a shallow aquifer and the shale formation is too large to make the upward fluid migration into the aquifer happen in a short time-scale [4, 31, 32]. Furthermore, shale layers are usually capped with tight formations acting as natural barriers [33]. Artificially created fractures may penetrate into the overburden formation, however, the distance to which they can propagate upward is limited and the fracture is very unlikely to pass through the overburden, reach the aquifer, and

288

Hydraulic Fracturing and Well Stimulation

connect it to the shale layer [31, 34, 35]. In addition, when the production starts fluid flows toward the horizontal section of the well, upward migration would be less probable to take place [32, 36]. Meanwhile, hydraulic fracturing operations near conductive faults might change the story and, as declared by Gassiat, Gleeson, Lefebvre, and McKenzie [37], under some circumstances, reservoir fluids could reach the aquifer through the permeable fault, although, due to the slow rate of movement, the mechanism takes place in a long time (thousands of years).

14.2 Risk Pathways to the Shallow Groundwater Risk pathways to shallow aquifers can be divided into two major categories: (1) over ground accidental spills from the pits used for temporary storage of the returned fluid or from accidents while transporting fluids or stemming from drilling activities (2) underground failure and upward fluid migration [30]. Upward fluid migration to the aquifer can take place through different pathways. USEPA suggests that some of the possible scenarios for the upward flow migration are as follows: 1. Damages in well integrity as a consequence of corrosion or improper cementing job. Leakage of fracking fluid into the formation can be a possible outcome of this scenario. 2. Poor design of the fracking job leading to fracture growth in overburden formations and providing hydraulic connections between the shale gas reservoir and aquifers above. In some cases, in presence of hydrocarbon zones above the shale layer, the situation becomes more critical as the fluid is capable of carrying hydrocarbon while traveling to shallower depths. 3. Fault reactivation resulted from the fracking process and provided permeable pathways for the fluid. 4. Fracturing the overburden formation and creating pathways into wells in the vicinity of the fracturing well (abandoned wells in most cases). In this situation, fluid might find its way through the aquifer using the adjacent wells and through issues in well integrity [3, 18]. Well integrity issues described in scenario 1 are further divided into two categories [28]: Behind the casing movement of the fluid toward the well head in which fluid finds the space between cementing and casing as a

Groundwater Contamination Potential

289

possible pathway (annular flow) and; leakage from the well in a radial pattern so that the fluid can reach the nearby formation and the chemicals can undergo different transport mechanisms (leak flow). King [1] has also named surface casing rupture as one of the potential leakage scenarios in a fracturing operation. Real-time pressure monitoring helps increase the chance of leakage discovery and, in a way, the pressure drop followed from a major leakage in pipe enables the operators to discover the leakage in five minutes [17]. However, when there is a smaller leakage event happening, the failure remains undetected during the hydraulic fracturing operation [17]. It is therefore, important to closely investigate the aftermath of a leakage event, particularly when the rupture is within the aquifer or at a point close to the aquifer. On the same subject, a probability bound analysis study has been done on risk of water pollution in Marcellus shale [30]. The study concluded that, among other risk pathways such as spills from on-site activities and wastewater disposal, well casing failure is an important contributor to the contamination risk where it shows lower than 0.01 m3 and 9 m3 contamination volumes for best-case 50th percentile and worst-case 50th percentile respectively [30]. Furthermore, a study by Browning and Smith [24] on deep well injection in Louisiana, Nebraska, Michigan, and Pennsylvania suggests malfunctioning casings and tubings as two reasons for gaining failure in mechanical integrity tests. According to the same study, out of 10,000 wells that had undergone the integrity test, 50% of them fail to pass with casing failure as the main contributor to the failure rate (responsible for 45% to 85% of failure). Among the failed tests due to casing problems, 22% showed failure in the only protective layer so that the injection waste was able to leak into the near formation. Although the data from this study might not be directly applicable to fracking wells because of the difference between waste injection and fracking fluid injection, it still gives ideas on chance of possible failures of casing in a fracking well [24].

14.3 Problem Statement This study is focused on the casing failure during the injection phase of hydraulic fracturing followed by a leakage flow into the aquifer. The fate of the contamination plume is investigated through application of a numerical solver for flow and transport equations. Various scenarios in this study are devised by altering hydrogeological parameters such as a homogeneous anisotropic permeability field and porosity, as well as operational parameters (i.e. leakage rate and depth of the leakage point). It is noteworthy to

290

Hydraulic Fracturing and Well Stimulation

state that the failure scenario of concern in this study falls into the type of incidents classified as low probability. The small probability is mainly attributed to the pressure monitoring during the operations and the fact that the leakage incident normally causes an observable pressure drop [17]. However, the leakage phenomenon is still worth studying, as the risk associated could be significant. One should note that risk, here, is defined as the multiplication of probability of an event by the consequence, and the aftermath of such a leakage scenario near or into the aquifer can be quite remarkable. With the low probability and high consequence, the hypothetical case studied here is considered to be a “rare” event. To investigate such a scenario a 2D numerical simulation is used and the effects of different parameters on the results are studied. The parameters of interest in this study are selected to be the concentration magnitude of the chemical and the arrival time at the well location.

14.4 Mathematical Formulation The chemical transport within the aquifer is modeled using a 2D flow field. The permeability field is set to be anisotropic homogenous and the porosity is kept constant. The transport equation is as follows [38]:

(c ) t

(c u

D c) q

(14.1)

in which c is the concentration of the chemical in the fluid phase, φ is the medium’s effective porosity, ρ is the water density, u is the specific discharge, and q is sink or source. The concentration varies in space x = (x,z) and time t. Darcy’s law for a fully saturated system is given by:

u

1

k( p

g z)

(14.2)

with k denoting an absolute permeability tensor and p, u representing pressure, and viscosity of water, respectively. The two dimensional diffusion and mechanical dispersion tensor is defined as follows:

D=

dmI |u| dl E(u ) dt E (u )

(14.3)

Groundwater Contamination Potential

291

where dm is the molecular diffusion coefficient and dl and dt are the longitudinal (parallel to the flow) and transverse (perpendicular to the flow in vertical direction) mechanical dispersion coefficients. The Euclidian norm of the specific discharge is given by:

|u|

u12 u22

(14.4)

The orthogonal projections in equation 3 are defined as:

E(u)

1 |u|2

u12 u1u2 u2u1 u22

u) and E (u ) I E(u

(14.5)

As for the boundary conditions used for the system of equations, we consider the Neumann type in which mass flux is prescribed on the boundary (Γ):

ρu · υ = g2 on Γ

(14.6)

where u is the unit normal outward to Γ and ρu·υ gives the projection of the flux vector on a unit normal vector of the boundary. In equation 14.6, g2 stands for the prescribed flux on the boundary. The initial condition is given as:

p(x,0) = p0(x), x

Ω

(14.7)

with p0(x) being the hydrostatic pressure and Ω denoting the entire domain of interest.

14.5 Hypothetical Case Description and the Numerical Method Figure 14.1 shows the leakage as a point source on the well casing. The layers are stacked on top of each other with the 20 meter under-saturated layer on top. The aquifer is also 20 meters thick and is separated from a much thicker sand formation through a 2 meter thick impervious layer. The 800 meter thick sand formation is playing a role as the overburden of the shallow shale

292

Hydraulic Fracturing and Well Stimulation

Injection well

Monitoring well Vadose zone Aquifer Impervious layer

Sand layer

Shale reservoir

Figure 14.1 Schematic representation of failure in vertical pipe and contamination plume in the aquifer.

layer extending from a depth of 840 to 900 meters. In order to investigate the effect of groundwater flow movement on the contamination, a monitoring well is placed 100 meters away from the injection well. The numbers here are adapted from average properties of shallow shale plays reported by the U.S. Energy Information Administration [39]. In the numerical analysis, the aquifer and the impervious layer are modeled using a two dimensional ECLIPSE grid with dimensions of 210 m × 20 m. ECLIPSE is a simulation software for subsurface multi-phase and multi-component flow and transport equations (see equations 14.1 through 14.7) [40]. The stochastic simulations are then conducted in order to find the probability of exceedance from the threshold in the monitoring well. The chemicals used in the fracturing fluid are modeled using a passive tracer and the results are reflected in measurements of one quantity of interest which is the concentration level at the monitoring well. The Monte-Carlo simulation is adapted to quantify the probability of crossing the specified concentration threshold. The fracturing operation is assumed to be carried out in 6 stages, each 2.5 hours long with 10 hours of relaxation time, with the injection rate of

Groundwater Contamination Potential

293

11,500 m3/d (50 bbl/ min). It is also presumed that the chemical concentration is equal to two weight percent of the fracturing slurry and the simulation time is 500 years. The parameters of concern for this study include porosity and anisotropic permeability of the aquifer, as well as the location of the leakage point with respect to the aquifer top and the leakage rate. Each parameter is randomly drawn from different probability distribution functions. Table 14.1 shows the parameters which are constant over the course of simulation. Stochastic simulations are conducted in order to evaluate the sensitivity of the model to different parameters and also to determine the probability of passing the contaminant threshold in the aquifer. The threshold is set to be 1 mg/l. Figure 14.2 shows the workflow used to incorporate randomness in the model. In here, the Monte-Carlo simulation is used to calculate the probabilities. The number of Monte-Carlo simulations and the probability Table 14.1 Parameters of the model. Parameter

Value

Aquifer Velocity

1.1 cm/d

Pressure Gradient in Aquifer

0.02 m/m

Tracer molecular diffusion coefficient

10-5 cm2/s

Tracer longitudinal dispersivity

0.5 m

Tracer transverse dispersivity

0.05 m

Tracer initial concentration

22 kg/m3

Value randomly drawn from specified probability density function

Simulation run

Saving the results

Figure 14.2 Random simulations workflow.

294

Hydraulic Fracturing and Well Stimulation

Table 14.2 Parameters of the random simulations. Parameter

Distribution

Mean

S.D.

Leakage rate (m3/d)

Uniform

1.05

0.55

Leakage depth (m)

Uniform

9.5

4.91

Poisson’s

18

18

Sand permeability (mD)

Log-normal

103.2

23.4

Sand porosity

Truncated Normal

0.175

0.04

Monte Carlo simulations number

1000

density functions (PDF) used for selecting the parameters are shown in Table 14.2. Log-normal and truncated normal distributions are chosen in a way to pick aquifer permeability and porosity values from ranges 150 to 270 mD and 0.10 to 0.30, respectively. Selected PDFs for permeability and porosity are in line with the petroleum engineering convention. The ranges chosen for permeability and porosity represent a permeable aquifer with an average permeability of 220 mD and an average porosity of 0.2. Also, the vertical permeability is selected as one-tenth of the horizontal permeability by convention. For the leakage location, two different PDFs are chosen. The first one is a uniform one extending from one to eighteen meters inside the aquifer with the assumption of absence of any information on the statistics of leakage on the wall of a vertical well. Whereas, the second is set to be a Poisson’s distribution in which the probabilities increase as the depth takes larger values in a manner where the maximum probability is gained for the depth equal to the aquifer bottom. The main reason for working with two different PDFs assigned for the leakage location is to assess the changes in results when one takes into account the fact that exertion of more stresses on the well casing in higher depths increases the probability of failure.

14.6 Results and Discussion The CDF of the normalized concentration at the monitoring well are shown in Figure 14.3 and Figure 14.4 for the cases of uniform and Poisson’s PDFs, respectively. For the first time-step shown (i.e. 0.27 years), the concentration plume is not yet reaching the monitoring well. At a time of 2.74 years, we can observe concentration values in the monitoring well for both cases of leakage PDF. The difference is that when working with the Poisson’s PDF,

Groundwater Contamination Potential 0.27 yr

0.5

00

1

3

2

4

0

27.4 yr

0.5

0 0

2

6

4 C/C0

1

2 C/C0

3 x 10–4

100 yr Concentration Threshold

1 CDF

CDF

0.5

5 x 10–5

C/C 0 1

2.74 yr

1 CDF

CDF

1

295

0.5

0

8 x 10–3

0

1

3

2

4

C/C0

5 x 10–5

Figure 14.3 Cumulative distribution functions (CDFs) of concentration value in the monitoring well – uniform distribution for the leakage location.

0.27 yr

2.74 yr 1 CDF

CDF

1

0.5

0

0

1

2

C/C0

3

4

0.5

0

5

6

–5

C/C0

x 10

27.4 yr

100 yr 1 CDF

CDF

1

0.5

00

2

4 0 C/C

8 x 10–4

6

8 x 10–3

Concentration Threshold

0.5

0

0

1

2

C/C0

3

4

5 x 10–5

Figure 14.4 Cumulative distribution functions (CDFs) of concentration value in the monitoring well – poisson’s distribution for the leakage location.

the concentration values show higher numbers as opposed to the case with uniform PDF. The probability of exceedance is also larger when using the Poisson’s PDF. The same holds for the next time-step shown (27.4 years), with a difference that the concentrations are roughly ten folds larger. In year 100 and beyond, the concentration values in the monitoring well show small numbers, mainly because of the mass transport from the open

296

Hydraulic Fracturing and Well Stimulation

Table 14.3 Probabilities calculated for different scenarios. P(C≥Cth)

Leakage Point PDF

Time (yr)

Uniform

Poisson’s

0.27

0

0

2.74

0.12

0.15

27.4

0.95

0.98

100

0

0

1 0.9 0.8

F(time)

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

2

4

6

8

(a)

10 Time (y)

12

14

16

1 0.9 0.8

F(time)

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 (b)

0

5

Time (y)

10

15

Figure 14.5 Arrival time CDF – Uniform (a) and Poisson’s (b) PDFs for the leakage depth.

Groundwater Contamination Potential

297

Table 14.4 Statistics of arrival time (years). Leakage point PDF

Mean

Variance

Min.

Max.

Range

Uniform

6.73

5.82

2.22

15.74

13.52

Poisson’s

5.57

2.98

1.7

14.74

13.04

boundary of the model. The calculated probabilities of passing the threshold value of the chemical scenarios are also listed in Table 14.3. In general, the probabilities take their maximum values at year 27.4 and show reduction from this time forward. Also, the case with Poisson’s PDF for the leakage depth shows slightly more critical results. The problem can also be viewed from the time perspective. The time of interest here (arrival time) is defined as the first moment after which the concentration level in the monitoring well goes beyond the specified threshold (1 mg/l). Figure 14.5 illustrates the CDF results for two cases of the leakage scenario (i.e. Uniform and Poisson’s PDF for the leakage depth) and Table 14.4 displays the statistics for different stochastic scenarios. On average, it takes between 6.73 and 5.57 years for the contaminant to go above its threshold in the monitoring well in this geological setting. It is also noticeable how replacing the uniform PDF by the Poisson’s one can affect the results.

14.7 Conclusion Hydraulic fracturing operations are not yet without deficiencies and can potentially be harmful to the environment if not practiced in a sound and safe manner. This work focuses on possible groundwater contamination issues stemmed from hypothetical failure and leakage in the well casing when the injection takes place. Permeability and porosity of the aquifer, leakage depth, and leakage rate are the parameters considered in this study. Higher importance is given to the leakage depth by assigning two different PDFs for this parameter and comparing the results. The results are reported as the concentration value of a tracer as well as the arrival-time of a specific threshold in the monitoring well. It can be inferred from the results that, under the special scenario and specific geological settings introduced in this study, the concentration in the monitoring well can exceed the threshold limit. It is also shown that using Poisson’s PDF results in more critical results as the concentrations reported increase and the average arrival-time decreases. Keeping the importance of leakage location in mind, one can

298

Hydraulic Fracturing and Well Stimulation

now impede failure scenarios by placing stronger obstacles in front of the fluid when the injection is happening. In general, some of the hydraulic fracturing drawbacks can be mitigated by setting stricter regulations and improving the design and well integrity. It is noteworthy to mention that a comprehensive study in a three-dimensional (3D) domain is required to give more solid ideas on the scenarios. 3D models help in achieving more realistic results, however, 2D simulations can still be helpful in understanding the general concepts. Also, one step forward in such studies could be increasing the uncertainty dimension by taking into account even more stochastic parameters, including the ones related to the physics of the well such as number of breaches on the well location. Those analyses require more advanced probabilistic methods to overcome the burden of the computation.

References 1. G.E. King, Hydraulic Fracturing 101: What Every Representative, Environmentalist, Regulator, Reporter, Investor, University Researcher, Neighbor and Engineer Should Know about Estimating Frac Risk and Improving Frac Performance in Unconventional Gas and Oil Wells, SPE Hydraulic Fracturing Technology Conference, The Woodlands, TX, pp. 1–80 (2012). 2. A. Daneshy, Hydraulic fracturing to improve production. SPE Tech. 101, 14–17 (1990). 3. USEPA (US Environmental Protection Agency), Study of the Potential Impacts of Hydraulic Fracturing on Drinking Water Resources, (2012). 4. J.D. Arthur, B.K. Bohm, B.J. Coughlin, M.A. Layne, and A L L Consulting, Evaluating Implications of Hydraulic Fracturing in Shale Gas Reservoirs, SPE Americas E&P Environmental and Safety Conference, pp. 23–25 (March 2009). 5. J. Heinecke, N. Jabbari, and N. Meshkati, The role of human factors considerations and safety culture in the safety of hydraulic fracturing (fracking). JSEE 2 (2), 130–151 (September 1, 2014). doi:10.7569/JSEE.2014.629509. 6. N. Jabbari, C. Ashayeri, and N. Meshkati, Leading Safety, Health, and Environmental Indicators in Hydraulic Fracturing, SPE Western Regional Meeting, Garden Grove, CA (2015). 7. T. Engelder, R.W. Howarth, and A. Ingraffea, Should fracking stop? Nature 477 (7364), 271–275 (2011). 8. L.M. McKenzie, R.Z. Witter, L.S. Newman, and J.L. Adgate, Human health risk assessment of air emissions from development of unconventional natural gas resources. Sci. Total Environ. 424, 79–87 (2012). 9. D.M. Kargbo, R.G. Wilhelm, and D.J. Campbell, Natural gas plays in the Marcellus shale: challenges and potential opportunities. Environ. Sci. Technol. 44 (15), 5679–5684 (August 1, 2010). doi:10.1021/es903811p.

Groundwater Contamination Potential

299

10. C.W. Moore, B. Zielinska, G. Pétron, and R.B. Jackson, Air impacts of increased natural gas acquisition, processing, and use: a critical review. Environ. Sci. Technol. 48 (15), 8349–8359 (March 18, 2014). doi:10.1021/ es4053472. 11. W.L. Ellsworth, Injection-induced earthquakes. Science 341 (6142), 1225942 (2013). 12. C.J. De Pater, and S. Baisch, Geomechanical Study of Bowland Shale Seismicity, Synthesis Report, 57 (2011). 13. R. Walker, F. Aminzadeh, and A. Tiwari, Correlation between Induced Seismic Events and Hydraulic Fracturing Activities in California, AGU Fall Meeting Abstracts, Volume 1, p. 05 (December 2014). 14. T. Göbel, A comparison of seismicity rates and fluid-injection operations in Oklahoma and California: implications for crustal stresses. The Leading Edge 34 (6), 640–648 (2015). 15. J. Healy, and W. Rubey, The denver earthquakes. Science 161 (3848), 1301– 1310 (1968). 16. ODNR (Ohio Department of Natural resources), Preliminary Report on the Northstar 1 Class II Injection Well and the Seismic Events in the Youngstown, Ohio, Area, (2012). 17. B.C. Gordalla, U. Ewers, and F.H. Frimmel, Hydraulic fracturing: a toxicological threat for groundwater and drinking-water? Environ. Earth Sci. 70 (8), 3875–3893 (August 13, 2013). doi:10.1007/s12665-013-2672-9. 18. USEPA (US Environmental Protection Agency), Assessment of the Potential Impacts of Hydraulic Fracturing for Oil and Gas on Drinking Water Resources, (2015). 19. R. Hammer, and J. VanBriesen, In fracking’s wake: new rules are needed to protect our health and environment from contaminated wastewater. Nat. Resour. Defense Coun. 11 (May 2012). 20. Y.R. Glazer, J.B. Kjellsson, K.T. Sanders, and M.E. Webber, Potential for using energy from flared gas for on-site hydraulic fracturing wastewater treatment in Texas. Environ. Sci. Technol. Lett. 1 (7), 300–304 (2014). doi:10.1021/ ez500129a. 21. J.M. VanBriesen, J.M. Wilson, and Y. Wang, Management of Produced Water in Pennsylvania: 2010-2012, Shale Energy Engineering, 107–113 (2014). 22. J.M. Nordbotten, M.A. Celia, S. Bachu, and H.K. Dahle, Semianalytical solution for CO2 leakage through an abandoned well. Environ. Sci. Technol. 39 (2), 602–611 (2005). doi:10.1021/es035338i. 23. C.F. Tsang, J. Birkholzer, and J. Rutqvist, A comparative review of hydrologic issues involved in geologic storage of CO2 and injection disposal of liquid waste. Environ. Geol. 54 (8), 1723–1737 (August 4, 2007). doi:10.1007/ s00254-007-0949-6. 24. L.A. Browning, and J.D. Smith, Analysis of the Rate of and Reasons for Injection Well Mechanical Integrity Test Failure, SPE/EPA Exploration and Production Environmental Conference, San Antonio, TX (1993).

300

Hydraulic Fracturing and Well Stimulation

25. R.J. Davies, S. Almond, R.S. Ward, R.B. Jackson, C. Adams, F. Worrall, L.G. Herringshaw, J.G. Gluyas, and M.A. Whitehead, Oil and gas wells and their integrity: implications for shale and unconventional resource exploitation. Mar. Petrol. Geol. 56, 239–254 (March 2014). doi:10.1016/j. marpetgeo.2014.03.001. 26. G.E. King, and D.E. King, Environmental risk arising from well-construction failure-differences between barrier and well failure and estimates of failure frequency across common well types locations. SPE Prod. Oper. 28 (4), 323– 344 (October 2013). 27. T. Watson, and S. Bachu, Evaluation of the potential for gas and CO2 leakage along wellbores. SPE Drill. Completion. 24 (1), 115–126 (March 2009). 28. M.D. Holloway, and O. Rudd, Fracking: The Operations and Environmental Consequences of Hydraulic Fracturing, John Wiley & Sons, Beverly, MA (2013). 29. N. Jabbari, F. Aminzadeh, F. de Barros, and B. Jafarpour, Hydraulic Fracturing and Potential Groundwater Contamination Risk. Presented at the 248th American Chemical Society National Meeting & Exposition, Paper Number: 315, San Francisco, CA, 10–14 (2014). 30. D.J. Rozell, and S.J. Reaven, Water pollution risk associated with natural gas extraction from the Marcellus shale. Risk Anal. 32 (8), 1382–93 (August 2012). doi:10.1111/j.1539-6924.2011.01757.x. 31. M. Zoback, S. Kitasei, and B. Copithorne, Addressing the Environmental Risks from Shale Gas Development, Worldwatch Institute, Washington, DC (2010). 32. R.W. Howarth, A. Ingraffea, and T. Engelder, Natural gas: should fracking stop? Nature 477 (7364), 271–275 (September 15, 2011). doi:10.1038/477271a. 33. J.D. Arthur, B.K. Bohm, and M.A. Layne, Hydraulic Fracturing Considerations for Natural Gas Wells of the Marcellus Shale, The Ground Water Protection Council Annual Forum, Cincinnati, OH (2008). 34. M. Fisher, and N. Warpinski, Hydraulic-fracture-height growth: real data. SPE Prod. Oper. 27 (1), 8–19 (November 2012). 35. S.A. Flewelling, M.P. Tymchak, and N. Warpinski, Hydraulic fracture height limits and fault interactions in tight oil and gas formations. Geophys. Res. Lett. 40 (14), 3602–3606 (July 28, 2013). doi:10.1002/grl.50707. 36. A. Kissinger, R. Helmig, A. Ebigbo, H. Class, T. Lange, M. Sauter, M. Heitfeld, J. Klünker, and W. Jahnke, Hydraulic fracturing in unconventional gas reservoirs: risks in the geological system, part 2. Environ. Earth Sci. 70 (8), 3855– 3873 (June 22, 2013). doi:10.1007/ s12665-013-2578-6. 37. C. Gassiat, T. Gleeson, R. Lefebvre, and J. McKenzie, Hydraulic fracturing in faulted sedimentary basins: numerical simulation of potential contamination of shallow aquifers over long time scales. Water Resour. Res. 49 (2) (December 12, 2013). doi:10.1002/2013WR014287.

Groundwater Contamination Potential

301

38. Z. Chen, G. Huan, and Y. Ma, Computational Methods for Multiphase Flows in Porous Media (Computational Science and Engineering), Society for Industrial and Applied Mathematic, Philadelphia, PA (2006). 39. USEIA (U.S. Energy Information Administration), Review of Emerging Resources: U.S. Shale Gas and Shale Oil Plays, US Energy Information Administration, (2011). 40. Schlumberger, ECLIPSE: Reservoir Simulation Software, Reference Manual, (2010).

Index Additives common fracturing, 58–60 manufactured ceramics proppants, 65–66 percentages of commonly used, 60–66 proppants, 61–63 resin coated proppant, 65 silica sand, 63–65 type and amount, 58 Air pollution, 18, 68, 244, 286 Alberta Geologic Survey, 168 Alfred P. Murrah Federal Building, 60 American Petroleum Institute (API), 15 Ammonium nitrate, 60 Anisotropy, 200 Anti-fracturing, 54 Appearance, 75 Armstrong PA, leakage scenario, 88–91 Artificial intelligence, 109–110 Artificial neural nets, 110, 111, 114 Auto-ignition temperature, 75 Back-propagation neural network model, 107–108 Base reservoir simulation model, 143–146 Bauxite, 65–66 Bessel function, 230–231 Big data, 94 Biocides, 59 Biot coefficient, 40

Biot modulus, 41 Boiling point, 74 Bottomhole pressure (BHP), 11, 39, 146, 184–185 Bowland Shale formation, 286 Bradford field, 87–88 Breaker solution, 59 Bulk modulus, 41 B-value analysis, 25–26, 106, 160, 188, 191–193 California Council on Science and Technology (CCST) report, 5, 14–15, 18 California Well stimulation technologies, 14–15 Carbon dioxide (CO2) emissions, 8 Care, in neural network design, 112 Cartesian coordinate system, 85 Cauchy’s formula, 46 CDFs. See Cumulative distribution functions (CDFs) Chemicals interaction of fracturing fluids, 277–278 manufacturers/importers, 72 storage, 70–72 Clean Air Act, 16 Clean Water Act, 16 Clustering analysis available data, 131–134 Fuzzy C-Mean algorithm, 130–131 Lost Hills oil field, 126–127 methodology and analysis, 131

Fred Aminzadeh (ed.) Hydraulic Fracturing and Well Stimulation, (303–313) © 2019 Scrivener Publishing LLC

303

304

Index

objective and motivation, 127–128 results and discussion, 136–139 technology, 128–129 CMG-BUILDER, 144–145 Coal, electricity generation from, 8 Cobra jet/max fracs, 127 Coherency-based stacking methods, 9 Cohesive element approach, 41 “Commitment to Excellence in HF,” 15 Completion cost of, 153–154 engineered completion approach, 128 parameters, 151–155 real time completion analysis technique, 182 Composite maps, 21 Computational efficiency, 82 Computational Fluid Dynamics (CFD) Simulation, 273 Constant parameters, stimulated reservoir volume, 209 Constraints, stimulated reservoir volume, 209–210 Continuous fracture modeling (CFM) technologies, 164 Corrosion inhibitors, 58, 59 Cost of completion, reducing, 153–154 Coulomb stress change, 169, 171, 174 CRAMP, 171 Crosslinking agent, 59 Crude oil average breakeven cost of, 28 production, 5 rise in cost of, 7 Cumulative distribution functions (CDFs) leakage scenario, 297 of normalized concentration, 294–295 Curable resin coated proppant, 65 Darcy’s units, 230 Darcy velocity, 41

Data acquisition, 85–86 Database maps, 21 Data conditioning, 85–86 Dayton Incident Well, 90 Decision variables, stimulated reservoir volume, 208 Deep resistivity log (DRES), 131 Deep-well injection process, 286 Density-based spatial clustering of applications with noise (DBSCAN), 130 Department of Energy (DOE), 7, 20 Depth-dependent lithology, 84 Devonian shale basins, 7 Discrete fracture network (DFN), 11 Division of Oil, Gas, and Geothermal Resources (DOGGR), 21 Drilling hydrocarbon, 80 operations, 66–68 rig components, 256 rig floor hazardous area layout, 256 safety culture, 254–257 Drilling induced fractures (DIFs), 80–81 Drill rigs, 68 Dynamic Finite Element Methods, 170 Earthquakes Alberta, 166–168, 173–178 in California, 20–21 caused by induced seismicity, 22 Oklahoma, 166–168, 173–178 potential using numerical material models, 168–173 ECLIPSE software, 292 Economics constraints, 218 factors, 29, 142 of hydraulic fracturing, 27–28 inter-lateral well spacing, 146 shale gas, 8 Utica-Point Pleasant shale play, 146 Edward-Trinity Plateau aquifer, 17

Index 305 Effective stimulated volume (ESV), 13 Elastic effects, 38 Embedment, 62–63 Emergency Planning and Community Right to Know Act, 16 Emission tomography, 9 Energy-based approach, 13 Energy Information Administration (EIA), 5, 248 Energy Information Agency, 27 Energy Policy Act of 2005, 94 Energy resources extraction, 28 fossil, 5 sustainable, 243 Engineered completion approach, 128 Environmental impacts, 14–18, 54–55, 243–245 Equivalent Fracture Models (EFM), 174, 177 Ergonomics, 259–261 Euclidean distance, 83 Evaporation rate, 75 Expectation-Maximization (EM) algorithms, 130 Extended sets, stimulated reservoir volume, 208–209 Extinguishing media, 76 Exxon Mobil subsidiary spills wastewater into the Susquehanna River, 264 Eye protection, 77 Fault stability, 160, 169 Fault zones, 21 Federal and State regulatory agencies, 16 Finite element method, 26 Fire & explosion hazard data, 75–76 Fire-fighting procedures, 76 Flammable limits, 75–76 Flashpoint, 75 Flowback calculated values, 278

causes of, 272 chemical interaction of fracturing fluids, 277–278 closing of fractures, 277 depth of penetration, 276–277 impact of parameters under control, 278–279 incremental oil production, loss in, 279–280 limitations, 281 reasons for partial, 275–278 Flowrate of well, calculations for, 283 Fluid injection, 257–258 Fluid mass balance equation, 40, 43 Fluids, 58 Fossil fuels, 243 Frac fluids, 4, 16, 258 calculations for damage, 282 chemical interaction of, 277–278 and composition, 57 uses and needs for, 57–58 FracFocus, 21 Frac stages, 128 Fracture contact problem, 43 Fracture geometry, 203–204 Fracture network, 80–83, 200–201 graph-based spatial analysis, 83–87 real world applications, 87–91 Fracture spacing, 101–103 applicability considerations, 120–121 artificial intelligence, 109–110 data, 110–114 method, 103–104 model fine-tuning, 108–109 natural fractures, 104–107 results, 114–120 workflow, 107–108 Fracture Zone Identifier (FZI) attributes, 10 Fracturing rocks, 7 Frequently Asked Questions (FAQ), 18 Friction reducers, 59

306

Index

Friction stress, 42 Fuzzy classification technique, 108–109 Fuzzy C-Mean (FCM) algorithm, 130–131, 134–136 Fuzzy Inference System, 110 Gamma ray (GR), 103, 132 Gas compressors, 67 Gas lift, 67–68 Gas resources, extraction, 27 Geertsma-De Klerk (KGD) models, 275 Gelling agents, 59 Geographic information system (GIS), 82 Geologic data, 84–85 Geomechanics evaluation, 11–14 of hydraulic fracturing, 11–14 microseismic mapping, 26 proxy, 174–175 regional faults, 166–168 Geometric interpretation of parameters, 221–225 stimulated reservoir volume, 203–204 Global Centroid Moment Tensor (CMT), 186–187 Graphs, 82–83 Graph-based approaches, 87–88 Graph-based spatial analysis, 83–87 Graph-based “topological” model, 83 Graph theory, 82 Greenhouse gas emission, 243 Griffth failure, 42 Groundwater contamination, 14–17 hypothetical case description, 291–294 mathematical formulation, 290–291 numerical method, 291–294 parameters of model, 293 probabilities calculation, 296 problem statement, 289–290

random simulations workflow, 293, 294 results and discussion, 294–297 shallow groundwater, 288–289 statistics of arrival time, 297 subsurface injection operations, 286 upward fluid migration, 287 Gutenberg-Richter law, 104 Hazard Communication Standard (HCS), 72 Hazard maps, 14 Hazardous decomposition products, 76 Hazardous polymerization, 76 Health hazard data, 76 HF. See Hydraulic fracturing (HF) HiWAY, 274 Holistic design techniques, 102 Horizontal drilling, 7 Human Error Probability (HEP), 260 Human factors considerations, 252–253 and safety culture, 259–261 Human, Organizational, and Technological (HOT) subsystems, 261–262 Hybrid microseismic survey design, 12 Hybrid neural network workflow, 103 Hydrafrac, 7 Hydraulically induced fractures (HIFs), 80–81 Hydraulic fracturing (HF) accidents, 263–265 additional types, 66 additives, common fracturing, 58–60 additives, percentages, 60–66 benefits of, 250 characterization, 8–10 chemical interaction of, 277–278 chemicals and products on locations, 70–72 closing of, 277

Index 307 definition, 4 diagnostics, 182, 188 disposal process, 20 drilling, 66–68, 254–257 earthquakes in California, 20–21 economics of, 27–28 energy transfer from, 49 environmental impacts, 14–18, 28, 54–55, 243–245 example, 4, 5 fault activation from, 20 field and laboratory setting, 16 fluid injection, 257–258 frac fluids, 57–58 fracking fluid, 258 geomechanics of, 11–14 horizontal and vertical stress, 8, 9 on horizontal/directional wells, 4 human factors considerations, 252–253 importance, 5–8 initiation and cessation, 11–12 intersection constraints, 224 mathematical formulation, 40–43 modelling, 275–276 on natural fractures, 46–49 numerical model, 43–44 oil and gas resources extraction via, 27 physical model, 39–40 pressurization, 48 risk factors, 16, 29 simulation results, 44–45 stage nodes representation, 223 stage parameters, 224 time evolution of effective stresses, 46–47 Tower of Babel, 55–57 upgraded visualization of, 273–275, 282 valid paths for neighboring wells, 224 wastewater, 258–259 Hydraulic fracturing Test Site, 11, 16

Hydraulic Fracturing Test Site I industry consortium (HFTS-1), 159–160 Hydrocarbon drilling, 80 Hydrochloric acid, 58–59 Hydrogen, 7 Illinois Basins, 7 IMEX, 144 Incompatibility, 76 Induced seismicity, 14–18 causes, 19 definition, 19 documented cases of injuries, 22–23 earthquake caused by, 22 in field and laboratory setting, 17 hydraulic fracturing causing, 19–20 in California, case study, 23–27 natural vs., 20 new workflow for estimating, 173–178 property damage caused by, 22–23 from wastewater injection, 21–22 Induced Seismicity Consortium (ISC), 20–21 Induced Seismicity Map (ISM), 24 Induced Seismicity Potential (ISP), 160, 172 in Alberta, 178 geomechanical proxy for, 174–175 injection volumes, 177 Induced Seismicity Potential in Energy Technologies Report, 22 Industry, and regulators, 93 Inflow equation, refractured oil wells, 235–241 Injuries, documented cases of, 22–23 In situ reservoir properties, 11 In situ stress, 11, 12 In-situ stress anisotropy, 38 Instantaneous event, 38 Integer Programming (IP), 161, 203

308

Index

Inter-lateral well spacing, 142 base reservoir simulation model, 143–146 capital and operating costs, 151 completion parameters, 151–155 cost of completion, 153–154 economic analysis inputs, 146 field cumulative gas production, 150 multi-lateral depletion, 148–151 net present value optimization study, 150 optimization scenarios, 147–148 single well case, 148 Iron control, 59 Kaolin, 65–66 k-mean clustering, 130 k-nearest neighbor algorithm, 83, 86 Armstrong PA, 90 Bradford field, 88, 89 connected graph, 92 on Dayton PA area, 92 wellbores, 91–93 Kronecker delta, 186 k sensitivity, 87–88 Landscape, changes in, 68–69 Large scale predictive model, 178–179 Leakage scenario, 88–91 Length-to-width ratio, 105 Lethal dose 50, 76 Linear elastic fracture mechanics (LEFM) model, 11 Linear momentum balance equations, 40 Linear programming-based approach, 202 Lithology, 84–85 Lost Hills oil field clustering analysis, 129–130 location, 126 objective and motivation, 127–128 technology, 128–129 Lower explosive limit (LEL), 75–76 Lower Marcellus Layer, 112

Macroergonomics, 260–261 Magma, 54 Manufactured ceramics proppants, 65–66 Marcellus play, 104 Mass balance equation, 40 Material Point Method (MPM), 160, 164, 170–173, 175, 178–179 Material Safety Data Sheets (MSDS), 72–73 contents of, 73 fire & explosion hazard data, 75–76 hazardous ingredients of mixtures, 74 health hazard data, 76 personal protection information, 77 physical data, 74–75 product identification, 73–74 reactivity data, 76 Matlab program, 83, 85, 86 Mechanical equilibrium equations, 40 Melting point, 74 Methane, 54 emissions, 18 leakage, 14 Michigan Basins, 7 Microearthquakes, 20 Microergonomics, 260 Microseismicity data analysis, 186–187 distribution, 11 of natural fractures, 48 Microseismic mapping extracting fracture geometry information, 9–10 geomechanical approach for, 26 stimulated reservoir volume, 201 Microseismic monitoring, 204 data, 187–188 design based on observations, 102 evaluation of, 12 pitfalls in analysis, 196 results, 188–195 Microseismic waveform, 13

Index 309 Midland Basin, 16 Model fine-tuning, 108–109 Modern Shale Gas Development in the United States, 61 Mohr-Coulomb theory, 42 Moment tensor inversion (MTI), 13 Monkeyboard, 255 Monte-Carlo simulation, 292–293 MSDS. See Material Safety Data Sheets (MSDS) Mud logs, 103, 107, 111–112 Mud spill in Pennsylvania state forest, 265 Multi-lateral depletion, 148–151 Multi-stage coiled tubing fracturing technique, 127 Multistage hydraulic fracturing, 101–102 applicability considerations, 120–121 artificial intelligence, 109–110 data, 110–114 method, 103–104 model fine-tuning, 108–109 natural fractures, impact of, 104–107 pilot wells, 103 results, 114–120 workflow, 107–108 National Academies of Science (NAS), 19 National Energy Technology Laboratory (NETL), 27, 88 National Energy Technology Laboratory Research Participation Program, 94 National Environmental Policy Act, 16 National Institute of Occupational Safety and Health (NIOSH), 249 National seismic hazard maps, 18 Natural fractures, 38 hydraulic fracturing effect on, 46–49 microseismicity of, 48

multistage hydraulic fracturing, 104–107 pre-existing, 39 Natural gas, 7–8, 54, 243, 248 Network training, 112 Neural network design, 112–114 Neuro-Fuzzy workflow, 103 Neutron porosity (NPHI), 132 Nitromethane, 60 Noise, 67–68 Nolte-Smith approach, 182–185, 188 Non-uniform time stepping, 44 Numerical material models, 168–173 Occupational Safety and Health Administration (OSHA), 72 Executive Summary, 72 permissible exposure limit, 74 Oda’s method, 142 Odor, 75 Oil extraction, 27 production, loss in incremental, 279–280 One-size-fits-all approach, 22–23, 102 One-way coupled approach, 42 Optimization procedure, 211–212 process, 205, 206 scenarios, 147–148 Organization of the Petroleum Exporting Countries (OPEC), 28 Overpressure, 44–45 Oxygen scavenger, 59 Pads, 68 Pennsylvania Department of Environmental Protection (PADEP) report, 90 Percentage volatile by volume, 75 Perkins-Kern-Nordgen (PKN) models, 275 Permeability log (PERM), 132 Permian Basin, 11

310

Index

Permissible exposure limit (PEL), 74 Personal protection information, 77 Petroleum extraction industry, 166 Plane stress model, 39 Point load, 62 Poisson’s ratio, 41, 107, 112 Porosity log (PHIT), 132 Prague fault, 176 Pre-cured resin coated proppant, 65 Pre-existing fractures, 11, 38, 39 Pressure barrier, 228 bottom hole, 11, 39, 146 data, 181 data analysis, 182–186 overpressure, 44–45 real-time pressure monitoring, 289 sinks, 228 vapor pressure, 74–75 Probabilistic-based simulations, 81 Probability density functions (PDF), 293–297 Production performance, 133 Product storage, 70–72 Property damage, caused by induced seismicity, 22–23 Proppants, 59, 61–63 coverage at end of pumping, 138 hardness, 63 manufactured ceramics, 65–66 resin coated, 65 shape, 63 silica sand, 63–65 size and placement, 62 Proxy for Coulombs Stress (PCS), 171–172 Pseudo-steady state flow conditions, 235–241 PyLith, 43 Qualitative methodology, 23 Quanta, 85–86 Quasi-static mechanical equilibrium, 40, 43

Rate transient analysis (RTA), 142 Reactivity data, 76 Real time completion analysis technique, 182 Real-time pressure monitoring, 289 Recovery factor (RF), 149, 151 Recycling wastewater, 29 Reduced-order models, 38 Refractured oil wells, 28 fracture conductivity on, 233 inflow equation, 235–241 mathematical model, 229–231 matrix-fracture cross flow, 229 model verification, 231 number of fractures on, 233 pressure barrier, 228 pressure sinks, 228 radial fractures, 229 reservoir and fracture properties, 232 secondary orthogonal fracture, 228 sensitivity analysis, 231–234 Regional bounding lithology, 84–85 Regional faults, 166–168 Regional-scale geological databases, 84 Regulators, industry and, 93 Research and development (R&D), 16 Reservoir volume, 8. See also Stimulated reservoir volume (SRV) Resin coated proppant, 65 Resource Conservation and Recovery Act, 16 Respiratory protection, 77 “Right to know” program, 71 Rocky Mountain Arsenal waste injection site, 244, 251, 287 Ryder trucks, 60 Safe Drinking Water Act, 16 Safety culture accidents, 263–265 common criticisms, 250–252 considerations, 261–263

Index 311 drilling, 254–257 Exxon Mobil subsidiary spills wastewater into the Susquehanna River, 264 fluid injection, 257–258 fracking fluid, 258 human factors considerations, 252–253 Mud spill in Pennsylvania state forest, 265 poor decision-making by management, 264–265 wastewater, 258–259 San Joaquin Valley, seismicity in, 23–27 SARA reporting, 71 Scale inhibitor, 59 Semi-analytical approach, 161, 231 Sets, stimulated reservoir volume, 208 Shale 2.0, 164 Shale gas, 7 economic impact of, 8 production of, 27 Shale oil, 5–7, 28, 248 Shale reservoir, 200–203 flow chart, 204 geometric interpretation, 203–204 mathematical model, 204–212 model building, 212–216 recommendations, 216–218 well and fracture design vector components, 204 Shale resources, 7 characterization, 27 development, 164 economics of, 8 exploitation, 202 North American, 166 regulations, 16 Shale rock, 200, 201, 248 Shale volume (VSH), 133 Shallow groundwater, risk pathways to, 288–289 Shear stress, 9, 12, 45, 172

Shear traction magnitude, 42 Short term exposure limit (STEL), 74 Signal-to-noise ratio (SNR) emissions, 9, 13 Silica sand, 63–65 Silicosis, 64–65 Siloed professionals, 173 Sintering process, 66 Skin protection, 77 Slick-water fracture, 7 Social license to operate (SLO), 165–166 Soil pollution, 68 Solubility in water, 75 Southern California Seismic Network (SCSN), 24 Sparse matrix, 85–86 Spatial analysis, 83–87 Specific gravity, 75 Spill Prevention Countermeasure and Control (SPCC) plans, 71 SRV. See Stimulated reservoir volume (SRV) Stability, 76 Stabilizing agents, 59 Stage spacing, 102 Stanolind Oil and Gas Corporation, 7 State regulators, 29 Stimulated reservoir volume (SRV), 48, 142, 161 assumptions and constraints, 207–210 complexity, 200 developed model flow chart, 204 discretized for one fracture stage, 225 drainage volume for, 211 equation, 201 geometric interpretation, 203–204 input parameters, 215 linear programming-based approach, 202 mathematical model, 204–207 methodology, 207

312

Index

microseismic mapping, 201 model building, 212–214 multigrid based, 203 network connections, 222 objective function, 207 for one fracture stage, 222 optimization procedure, 205, 206, 211–212 output parameters, 215–216 overlapping design, 207 physical and economic constraints, 218 properties of reservoir model, 217 recommendations, 216–218 representation, 210–211 results, 216 size of, 200 staggered design, 207 well and fracture design vector components, 204 well pad and evaluation, 214–216 Streamtube model assumptions, 272 limitations, 281 upgraded visualization of hydraulic fracturing, 273–275, 282 Stresses, 84 field, 11, 45, 160, 164, 169, 172, 174 friction, 42 plane stress model, 39 shadowing, 129 shear, 9 in situ stress, 11, 12 volumetric effective stress, 41 Stress tensor, 40 Subsurface Fluid Injection and Production (SFIP) strategy, 13 Subsurface geology, 84–85 Surfactant, 59, 60 Time weighted average (TWA), 74 Topological algorithm, 85–86 big data, 94

real world applications, 87–91 value of, 86–87 Tower of Babel, 55–57 Traditional geospatial models, 82, 86 Traffic problems, 69–70 True Vertical Depth (TVD), 24 Underground Injection Control (UIC) well data, 167 Undirected graph, 82 United States Environmental Protection Agency (USEPA), 287 Upper explosive limit (UEL), 75–76 Upscaled models, 38 U.S. Energy Information Administration (EIA), 5, 248 U.S. Geological Survey (USGS), 19–20 US Seismic Hazard Map, 18, 24 Utica-Point Pleasant shale play base reservoir simulation model, 143–146 capital and operating costs, 151 completion parameters, 151–155 cost of completion, 153–154 economic analysis inputs, 146 field cumulative gas production, 150 location, 143 multi-lateral depletion, 148–151 net present value optimization study, 150 optimization scenarios, 147–148 single well case, 148 Vapor density, 75 Vapor pressure, 74–75 Variables, stimulated reservoir volume, 208 Vector test function, 41 Velocity model, 13 Ventilation, 77 Viscosity, 75 Visual pollution, 68 Volumetric effective stress, 41 Volumetric strain, 41

Index 313 Wastewater, 258–259 disposal, 164, 168, 177, 289 injection, induced seismicity from, 21–22 recycling, 29 Waterflooding, 127, 131 Water pollution, 68, 287, 289 Water usage, amount of, 17–18 Weak solid-to-fluid coupling, 41 Wellbores, 81–82 k-nearest neighbor algorithm, 91–93 point data, 85

spatial locations of, 91 Well integrity, 287–289, 298 Well pad, 214–216 Well perforation design, 128, 257 Well placement process, 220–221 Well spacing, 142 Wolfcamp Delaware basin, 28 Young’s modulus, 41, 107, 171 Zero-displacement boundary condition, 12