«Modeling and Characterization of Hydraulic Stimulation and Induced Seismicity in Geothermal and Shale Gas Reservoirs Mark McClure December 2012 ...»
Modeling and Characterization of Hydraulic
Stimulation and Induced Seismicity in Geothermal and
Shale Gas Reservoirs
Financial support was provided through the
Precourt Institute of Energy
and by the Department of Energy Resources Engineering,
Stanford Geothermal Program
Interdisciplinary Research in
Engineering and Earth Sciences
Stanford, California Abstract The classical concept of hydraulic fracturing is that large, wing-shaped tensile fractures propagate away from the wellbore. However, in low matrix permeability settings such as Enhanced Geothermal Systems (EGS) and gas shale, hydraulic fracturing creates complex networks that may contain both newly formed fractures and stimulated natural fractures.
In this research, the overall approach has been to integrate field observations, laboratory observations, and understanding of fundamental physical processes into computational modeling that is specifically designed for complex hydraulic fracturing and to apply the modeling to develop deeper understanding and to solve practical problems.
A computational model was developed that coupled fluid flow, stresses induced by fracture opening and sliding, transmissivity coupling to deformation, friction evolution, and fracture propagation in two-dimensional discrete fracture networks. The model is efficient enough to simulate networks with thousands of fractures. A variety of novel techniques were developed to enable the model to be accurate, efficient, realistic, and convergent to discretization refinement in time and space. Testing demonstrated that simulation results are affected profoundly by the stresses induced by fracture deformation, justifying the considerable effort required to include these stresses in the model.
Four conceptual models were formulated that represent the main hypotheses about stimulation mechanism from the literature of hydraulic fracturing. We refer to the stimulation mechanisms as Pure Opening Mode (POM), Pure Shear Stimulation (PSS), iv Mixed-Mechanism Stimulation (MMS), and Primary Fracturing with Shear Stimulation Leakoff (PFSSL). Computational models were used to investigate the properties of each mechanism. Geological factors that affect stimulation mechanism were identified.
Techniques for diagnosing stimulation mechanism were devised that incorporate interpretation of bottom hole pressure during injection, shut-in, and production, microseismic relocations, and wellbore image logs. A Tendency to Shear Stimulation (TSS) test was proposed as a way to help diagnose the mechanism by unambiguously measuring a formation's ability to experience shear stimulation. Modeling results suggested several potential sources for error in estimation of the least principal stress in low matrix permeability settings. The Crack-like Shear Stimulation (CSS) mechanism was identified as a potentially important physical process that may control the spreading of shear stimulation through the interaction of fluid flow, deformation, and sliptransmissivity coupling.
The computational model also has the capability to couple fluid flow with rate and state earthquake simulation. The model was used to investigate the interaction of fluid flow, permeability evolution, and induced seismicity during injection into a single large fault. Using the model, a variety of observations about induced seismicity in EGS were explained. Producing fluid back after injection and gradually reducing injection pressure during stimulation were identified as strategies for minimizing induced seismicity.
A review of historical EGS projects demonstrated that the severity of induced seismicity has been correlated to the degree of brittle fault zone development in the interval of injection. The fracture networks at each project were categorized along a continuum from thick, porous fault zones to thin cracks. Observations from specific EGS projects fell across the full continuum, a result that has implications not only for induced seismicity, but for fractured reservoirs in general.
A pressure transient analysis was performed using data from the EGS project at Soultz-sous-Forêts. At Soultz, fluid injection induced slip and transmissivity enhancement in large fault zones. The pressure transient analysis showed that these fault zones are best described as slabs of single porosity, single permeability material.
Evidence of dual porosity behavior was not found.
Thank you to my advisor Dr. Roland Horne. Roland has been tirelessly supportive and always pushed to get the best out of me. Not only has Roland been a tremendous mentor for research, he has had a profoundly positive influence on the direction that my life has taken over the years. I do not think that you could ask for more from an advisor.
There have been many people who have been very generous to meet with me informally as I have pursued my PhD and let me tap into their knowledge and expertise.
First, thank you to Dr. Mark Zoback and Dr. David Pollard, who I met with particularly frequently and have been very influential. Also thank you to Dr. Eric Dunham, Dr. Paul Segall, and Dr. Hamdi Tchelepi.
Thank you to Dr. Pollard for providing the code COMP2DD, which was used to verify the accuracy of the model used in Chapter 2. Thank you to Dr. Dunham for providing an earthquake simulation code that was used to verify the accuracy of the model used in Chapter 4.
Dr. Andrew Bradley was influential in the development of the modeling described in Chapters 2, 3, and 4. Most importantly, his matrix multiplication code Hmmvp is integrated into the simulator used in Chapters 2 and 3, and it plays a crucial role in making the simulator efficient. Not only was he very generous to share his code, he modified it so that it could be better integrated with the simulator, and he spent time helping me to implement it.
Thank you to Dr. Albert Genter and Dr. Nicolas Cuenot for supplying the data from the Soultz EGS project that was used in Chapter 6. In gathering information for Chapter 5, I contacted several people from around the world who supplied me with data, vi references, and/or stimulating discussion. Thank you to Dr. Nick Davatzes, Dr. Hideshi Kaieda, Dr. Ingrid Stober, Dr. Albert Genter (again), Dr. Keith Evans, Dr. Eva Schill, Dr.
Steve Hickman, Dr. Doone Wyborn, and Dr. Günter Zimmermann.
Thank you to the Precourt Institute for Energy at Stanford for providing the financial support that made this work possible.
Thank you to my friends for being supportive over the years that I have pursued my PhD. Often, I have said that I could not do this or that because I had to do research, and thanks for understanding. But also, thank you for making sure that I did not work too much! Thank you to my officemates Obi, Mike, and Charles for being such good companions over the past three years. Thank you to all of my fellow students and especially to my colleagues in the Stanford Geothermal Program, for all the good times and stimulating conversation.
Finally, thank you to my parents, grandparents, and siblings, who inspire me by their example.
Table of Contents
List of Tables
List of Figures
1.2 Modeling Philosophy
1.3 Scope of Work
2 Modeling Methodology, Validation, and Testing
2.1.1 Discrete Fracture Network Modeling
2.1.2 Review of Stimulation Models
2.2.1 Governing and Constitutive Equations
2.2.2 Initial Conditions
2.2.3 Methods of Solution
188.8.131.52 Iterative Coupling
184.108.40.206 Fracture Deformation: Displacement Discontinuity Method
220.127.116.11 Stresses Induced by Normal Displacements of Closed Fractures
18.104.22.168 Solution to the Fluid Flow and Normal Stress Equations
22.214.171.124 Solution to the Shear Stress Equations
126.96.36.199 Inequality Constraints on Fracture Deformations
188.8.131.52 Changing Mechanical Boundary Conditions
184.108.40.206 Formation of New Tensile Fractures
220.127.116.11 Adaptive Time Stepping
18.104.22.168 Wellbore Boundary Conditions
4 Spatial Domain
22.214.171.124 Generation of the Discrete Fracture Network
126.96.36.199 Spatial Discretization
2.2.5 Special Simulation Topics
188.8.131.52 Efficient Matrix Multiplication
184.108.40.206 Crack Tip Regions
220.127.116.11 Dynamic Friction Weakening
18.104.22.168 Alternative Methods for Modeling Friction
22.214.171.124 Adaptive Domain Adjustment
126.96.36.199 Strain Penalty Method
188.8.131.52 Neglecting Stresses Induced by Deformation
2.3.1 Simulation and Discretization Details
2.3.2 Model A: Small Test Problem
184.108.40.206 Solving Directly for the Final Deformations
220.127.116.11 Testing the Effect of cstress
2.3.3 Models B and C: Large Test Problems
18.104.22.168 Model B: Large Test Problem of Shear Stimulation
22.214.171.124 Model C: Large Test Problem of Mixed-Mode Stimulation
2.3.4 Model D: Testing the Strain Penalty Method
2.3.5 Hierarchical Matrix Decomposition
2.4.1 Model A
126.96.36.199 General Description of Results
188.8.131.52 Effect of Spatial and Temporal Discretization
184.108.40.206 Solving Directly for Final Deformations
220.127.116.11 Effect of cstress
2.4.2 Model B
18.104.22.168 General Description of Results
22.214.171.124 Effect of Temporal Discretization Refinement
126.96.36.199 Effect of cstress
188.8.131.52 Effect of Adaptive Domain Adjustment
184.108.40.206 Dynamic Friction Weakening
220.127.116.11 Neglecting Stress Interaction
2.4.3 Model C
2.4.4 Model D
2.4.5 Hierarchical Matrix Decomposition
2.4.6 Extension of the Model to Three Dimensions
ix 3 Conceptual Models of Stimulation
3.1.1 Mechanisms of Stimulation
3.1.2 Practical Consequences of Stimulation Mechanism
3.1.3 Effect of Geological Parameters on Stimulation Mechanism
18.104.22.168 Requirements for Shear Stimulation
22.214.171.124 Requirements for the PSS Mechanism
126.96.36.199 Fracture Termination
3.2.1 Details of the Stimulation Model
3.2.2 Details of the Simulations
3.4.1 Pure Opening Mode Fracturing
3.4.2 Pure Shear Stimulation
188.8.131.52 Prototype of Shear Stimulation
184.108.40.206 Initial Transmissivity and Crack-like Shear Stimulation
220.127.116.11 Shear Stimulation and Void Aperture
3.4.3 Mixed Mechanism
3.4.4 Primary Fractures with Shear Stimulation Leakoff
3.4.5 Bottom Hole Pressure During Injection and Shut-in
3.4.6 Production Behavior
3.4.7 Estimation of the Least Principal Stress
3.4.8 Testing a Formation's Tendency for Shear Stimulation
3.4.10 Wellbore Logs
3.4.11 Summary of Methods to Diagnose Stimulation Mechanism
4 Investigation of Injection-Induced Seismicity using a Coupled Fluid Flow and Rate/State Friction Model
4.1.2 Summary of Results
4.1.3 Seismicity Modeling in EGS
x 4.1.4 Relationship of our Model to Actual EGS Reservoirs
4.2.1 Problem Definition
4.2.2 Methods of Solution
4.2.3 Time Discretization
4.2.4 Problem Setup
4.4.1 Similarity and Differences Compared to EGS Field Observations
4.4.2 Episodic Crack-Like Shear Stimulation
4.4.3 Shut-in Events
4.4.4 Changes in Flow Rate with Time
4.4.5 Implications of the Crack-like Shear Stimulation Mechanism for Estimation of the Unstimulated Hydraulic Diffusivity
4.4.6 Implications of the Crack-like Shear Stimulation Mechanism for Estimation of the Least Principal Stress
4.4.7 Effect of Injection Pressure for Constant Pressure Injection
4.4.8 Effect of Decreasing Injection Pressure over Time
4.4.9 Effect of Producing Fluid Back after Injection
4.4.10 Effect of Shear-Induced Pore Volume Dilation
4.4.11 Effect of dc
4.4.12 Comparison of Rate/State Friction to Static/Dynamic Friction
5 The Effect of Fault Zone Development on Induced Seismicity................ 247 5.1 Introduction
5.1.2 Effect of Fault Development
5.1.3 Summary of Results
xi 5.3.3 Ogachi
5.3.5 Bad Urach
5.3.9 Groβ Schönebeck
5.3.10 Cooper Basin
5.4.1 Slip Surface Continuity
5.4.2 Seismic and Aseismic Slip
5.4.3 Fracture Orientation
5.4.4 Dependence on Depth
5.4.5 Alternative Mechanisms of Acoustic Emission
5.4.6 Background Seismicity
5.4.7 Outlier Events and Geological Heterogeneity
5.4.8 Magnitude-Frequency Distribution
5.4.9 Seismic Hazard Analysis
6 Pressure Transient Analysis of Fracture Zone Permeability at Soultz-sousForêts 285 6.1 Introduction
6.1.1 The Soultz Reservoir
6.1.2 Fracture Opening vs. Fracture Slip
6.1.3 Flow in the Unfractured Granite
6.1.4 EGS Modeling
6.1.5 Fault Zones
6.2 Pressure Transient Analysis of GPK2
6.2.2 Model Construction
6.2.3 Data Matching
7.1 Future Work
xiiiList of Tables
Table 2-1: Discretization settings
Table 2-2: Baseline settings for all simulations.
Table 2-3: Model specific baseline settings.
Table 2-4: Deviations from baseline settings for Simulations S0-S12.
Table 2-5: Deviations from baseline settings for Simulations B1-B9.
Table 2-6: Settings used for the discretizations in Figure 2-38.
Table 3-1: Summary of terminology used in Chapter 3