Evolution of Permeability and Induced Seismicity within Fractured Reservoirs

Open Access
Author:
Izadi, Ghazal
Graduate Program:
Energy and Mineral Engineering
Degree:
Doctor of Philosophy
Document Type:
Dissertation
Date of Defense:
August 08, 2012
Committee Members:
  • Derek Elsworth, Dissertation Advisor
  • Derek Elsworth, Committee Chair
  • Zuleima T Karpyn, Committee Member
  • Li Li, Committee Member
  • Chris Marone, Committee Member
  • Yilin Wang, Committee Member
Keywords:
  • permeability Evolution
  • Induced Seismicity
Abstract:
Fractured in geologic media, advective transport, heat transfer and chemical transport can change the porosity during short- to long-term fluid circulation that results from changes in mineral volume fractions. The porosity-permeability correlation in geologic media can be complex and depends on several factors such as fracture size distribution, fracture orientations, fracture network connectivity and in situ stresses, among others. Within fractured reservoirs under geothermal conditions, coupling between fluid transport, mechanical response, heat transfer and chemical reactions may change the permeability and influence the induced seismicity both during short-term reservoir stimulation and long-term production. During short-term stimulation, rapid enhancement of fracture permeability occurs as a result of circulating fluid through these fracture networks. During long-term production, hydraulic and thermal effects both contribute to the reactivation of the natural fracture networks and that also enhances the reservoir permeability. During the reactivation of pre-existing fractures due to hydraulic transport and mechanical deformation, seismic events occur when fluid circulates between the injection and production wells with a large number of high magnitude events associated with the initiation of the reservoir stimulation. Determining dominant behaviors that control the enhancement of permeability and the triggering of induced seismicity is the main focus of this work. This dissertation examines the role of coupled thermal, hydraulic, mechanical and chemical effects on various fracture networks in promoting the failure of pre-existing fractures and their influence on the evolution of seismicity and of permeability. Part I of this thesis (Chapters 1 and 2) explores the behavior of engineered geothermal systems under the response of complex interactions that change the permeability and induce seismicity. First a model is introduced to evaluate the rate of propagation of seismicity within a fractured reservoir that represents the Cooper Basin geothermal field. This model is used in Chapter 1 to explore the spatial and temporal distribution of seismic activity within reservoir and to replicate observed patterns. We then apply this understanding (Chapter 2) to more complex fracture networks utilized to explore the reservoir behaviors during reservoir stimulation to production and in understanding the effects that can promote induced seismicity. This model is applied for both the Cooper Basin and Coso geothermal field, for which data are available. The models developed and validated in Chapters 1 and 2 are then applied to project the response of the Newberry EGS demonstration project to both stimulation (Chapter 3) and long-term production (Chapter 4). Part II (Appendix) explores the role of strains developed in unconstrained sorbing and swelling fractured media (coal) and their role in evolving permeability and in understanding the behavior when CO2 is injected into ECBM or into sequestration sites. In chapter 1, we use a continuum model of reservoir evolution to explore the interaction of coupled thermal, hydraulic and chemical processes that influence the seismicity evolution from stimulation through production. Events occur from energy released of seeded fractures enabling moment magnitude, frequency and spatial distribution to be determined with time. We evaluated the magnitude of events which varies in a range of -2 to +2. The largest event size (~2) corresponds to the largest fracture size (~500m) and a prescribed stress drop of 9MPa. Modeled b-values (~0.6 to 0.7) correspond to observations (~0.7 to 0.8) and this approach is successfully calibrated against observations in the Cooper Basin (Australia). Then we tracked the hydrodynamic and thermal fronts to define causality in the triggering of seismicity. The hydrodynamic front moves twice as fast as the thermal front and envelops the triggered seismicity at early time (days to month) – with higher flow rates correlating with larger magnitude events. For later time (month to years) thermal drawdown and potentially chemical influences principally trigger the seismicity but result in a reduction in both the number of events and their magnitudes. In chapter 2, we explore the role of thermal, hydraulic and chemical effects on the evolution of seismicity in reservoirs from stimulation through production. We use a continuum model capable of accommodating changes in stress that result from change in fluid pressure as well as thermal stress. Discrete penny-shaped fractures are seeded within the reservoir volume and failure of fractures is calculated from the finite difference model FLAC3D. Energy release magnitude is utilized to obtain the magnitude-moment relation for induced seismicity both by location and with time. We calibrate fracture spacing and length of fractures in the model and replicate observed b-values during reservoir production (10years). We observe that the larger the energy release the larger the number of events induced at a given location and the greater the probability of large-magnitude events. As the seeded fracture size is increased the moment-magnitude also increases. Maximum magnitudes for identical fracture distributions are Ms ~2 for the Cooper Basin and ~1.2 for Coso corresponds to the largest fracture size (~500m) and a prescribed stress drop of 9MPa and 3MPa. We applied low and high (0.1m-1 vs. 0.2m-1) fracture densities in the model for Cooper Basin reservoir to explore the behavior of moment magnitude distribution. For the widely spaced fracture networks (0.1m-1), the increase in rate of propagation of events reaches a smaller radius from the injection point and induces seismicity evolution is slower with time compared to the behavior of the closely spaced fracture network (0.2m-1). We observed that the modeled b-values (~0.6 to 0.7) also correspond to observations (~0.7 to 0.8). Then we tracked the hydrodynamic and thermal fronts and to define causality in the triggering of seismicity. The hydrodynamic front moves twice as fast as the thermal front and envelops the triggered seismicity at early time (days to month) – with higher flow rates correlating with larger magnitude events. For later time (month to years) thermal drawdown and potentially chemical influences principally trigger the seismicity but result in a reduction in both the number of events and their magnitude. In chapter 3, we utilize a continuum model of reservoir behavior subject to coupled THMC (thermal, hydraulic, mechanical and chemical) processes to explore the evolution of stimulation-induced seismicity and of permeability in EGS reservoirs. Our continuum model is capable of accommodating changes in effective stresses that result due to the evolving spatial variations in fluid pressure as well as thermal stress and chemical effects. Discrete penny-shaped fractures (~10-1200m) are seeded within the reservoir volume at prescribed (faults) and random (fractures) orientations and with a Gaussian distribution of lengths and location. Failure is calculated from a continuum model using a Coulomb criterion for friction. Energy release magnitude is utilized to obtain the magnitude-moment relation for induced seismicity by location and with time. This model is applied to a single injector (stimulation) to the proposed Newberry EGS field (USA). We stimulate the reservoir in four zones of differing fracture network properties B, C, D and E (shallow to deep) and at four different depths of 2000, 2500, 2750 and 3000 m. The same network of large fractures (density of 0.003 m-1 and spacing 300 m) is applied in all zones and supplemented by more closely spaced fractures with densities of 0.5 m-1 in the shallow zone B, 0.9 m-1 in the intermediate zones C and D and 0.26 m-1 in the deepest zone E. We show that permeability enhancement is modulated by hydraulic, thermal, and chemical (THMC) processes and that permeability increases by an order of magnitude during stimulation at each depth. For the widely spaced fracture networks, the increase in permeability reaches a smaller radius from the injection point and permeability evolution is slower with time compared to the behavior of the closely spaced fracture network. For seismic events that develop with the stimulation, event magnitude (MS) varies from -2 to +1.9 and the largest event size (~1.9) corresponds to the largest fractures (~1200m) within the reservoir. We illustrate that the model with the highest fracture density generates both the most and the largest seismic events (MS =1.9) within the 21 day stimulation. Rate of hydraulic and thermal transport has a considerable influence on the frequency, location and time of failure and ultimately event rate. Thus the event rate is highest when the fracture network has the largest density (0.9m-1) and is located at depth where the initial stresses are highest (zone D). Apparent from these data is that the closely spaced fracture network with the higher stress regime (at the deeper level) has the largest b-value ~0.74. In chapter 4, we utilize a numerical model to examine thermal-hydrologic-mechanical-chemical processes leading to the evolution of induced seismicity in naturally fractured dual-porosity media. We use a continuum model to examine the thermo-hydro-mechanical behaviors due to variation in fluid pressure and thermal stress on different fracture networks within a prototypical enhanced geothermal system (EGS). Discrete penny-shaped fractures are seeded within the reservoir volume with random orientations and a Gaussian distribution of lengths. Failure is calculated from Coulomb strengths. Energy release magnitude is utilized to obtain the magnitude-moment relation for induced seismicity by location and with time. This model is applied to the potential Newberry EGS field (USA) by assuming fracture sizes of 10 to 1200 m. Models are classified by their conceptualization of the fractured reservoir geometry as both networks of discrete fractures and with equivalent fractured media as fill-in. This model is applied to a doublet injector-producer to explore the spatial and temporal triggering of seismicity for varied fracture network geometries at shallow (2000m) and deep (2750) depths. First we consider the identical network of large fractures (300 m fracture spacing) in both shallow and deep zones and infilled with smaller (10-200m) more closely spaced fractures with densities of 0.5 m-1 in the shallow zone (B) and of 0.9 m-1 in the deeper zone D. Then we apply a different network where the spacing of the large fractures are halved (~150m) in both zones but with the small closely spaced fractures retained with densities of 0.5m-1 in the shallow zone (B) and 0.9m-1 in the deeper zone (D). We evaluate the magnitude of seismic events that vary from -2 to +1.9 with the largest event size (~1.9) corresponding to the largest fracture size (~1200m) within the reservoir. We illustrate that the model with the higher fracture density generates both the most and the largest seismic events (MS =1.9), thus the evolution of seismicity is quickest and migration of seismic events is fastest with radius from the injector compared with the case for more widely-spaced fractures. Rate of hydraulic and thermal transport has a considerable influence on the amount, location and time of failure and ultimately event rate. Thus, the event rate is higher when the fracture network has the larger density (0.9m-1) with closely-spaced fractures (150m) and is located at depth where the initial stresses are highest (zone D). In addition, the modeled b-value shows that there is a relation between fracture networks, fracture spacing, and the evolution of seismicity. In all cases, the a-value is decreased and the b-value is increased with time and this indicates small-magnitude events with a small number of events induced during long term production. Apparent from these data is that the closely spaced fracture network with the higher fracture density and stress regime (at the deeper level) has the largest b-value ~1.34. Finally, we evaluate the thermal energy recovered during the production and the results show that the highest thermal energy is recovered from the deeper zone (D) with the more closely-spaced fractures (150m). In appendix, we explore the conundrum of how permeability of coal decreases with swelling-induced sorption of a sorbing gas, such as CO2. We show that for free swelling of an unconstrained homogeneous medium where free swelling scales with gas pressure then porosity must increase as pressure increases. The volume change is in the same sense as volume changes driven by effective stresses and hence permeability must increase with swelling. An alternative model is one where voids within a linear solid are surrounded by a damage zone. In the damage zone the Langmuir swelling coefficient decreases outwards from the wall and the modulus increases outwards from the wall. In each case this is presumed to result from micro-fracturing-induced damage occurring during formation of the cleats. We use this model to explore anticipated changes in porosity and permeability that accompany gas sorption under conditions of constant applied stress and for increments of applied gas pressure. This model replicates all important aspects of the observed evolution of permeability with pressure. As gas pressure is increased, permeability initially reduces as the material in the wall swells and this swelling is constrained by the far-field modulus. As the peak Langmuir strain is approached, the decrease in permeability halts and permeability increases linearly with pressure. This behavior is apparent even as the constraint on damage around is relaxed and ultimately removed to represent a homogeneous linear solid containing multiple interacting flaws. In either case the rate of permeability loss is controlled by crack geometry, the Langmuir swelling coefficient and the void “stiffness” and the rate of permeability increase is controlled by crack geometry and void “stiffness” alone. Permeability evolution may be approximated by a single non-dimensional variable incorporating fracture spacing, flaw-length, Langmuir strain and initial permeability. This model represents the principal features of permeability evolution in swelling media and is a mechanistically consistent and plausible model for behavior.