Investigation of Injection-induced Seismicity Using a Coupled Fluid Flow and Rate/state Friction Model

This paper describes a numerical investigation of seismicity induced during injection into a single, isolated fracture. A model was developed and used that couples (1) fluid flow and (2) rate and state friction. Rate and state friction theory describes how friction on a fault depends on both sliding velocity and past sliding history. Rate and state friction is used widely in earthquake modeling. We investigated the effect of various factors, both geological and practical, that impact induced seismicity. Our modeling indicated that shearinduced pore dilation may be an important process that prevents injection from triggering slip events that propagate far from the injector. The effect of injection schedule was examined. It was found that decreasing injection pressure over time was a successful strategy for reducing the maximum event magnitude. The model predicted that significant seismicity should occur after injection ceases, which is consistent with observations from EGS stimulations. The post-injection events were caused by a redistribution of fluid pressure after injection. Production of fluid from the well immediately after injection inhibited post-injection events. In prior work, we investigated the impact of injection schedule using a simpler treatment of friction that we refer to as static/dynamic. The injection schedule findings described in this paper were consistent with our results from our static/dynamic modeling. Despite the apparent success of the static/dynamic modeling, it has important drawbacks, and we discuss some of those issues in this paper. Our eventual goal is to perform coupled fluid flow and rate and state simulation on a large network of fractures. We discuss various strategies we have implemented or are implementing to improve the efficiency of the simulations and make large scale simulations possible. INTRODUCTION Enhanced Geothermal Systems (EGS) are characterized by the use of hydraulic stimulation to enhance flow rate in high temperature, low productivity wells, typically located in crystalline basement rock. Water is injected at high pressure without proppant, and in most cases the increased fluid pressure triggers slip on pre-existing fractures. The process of using increased fluid pressure to trigger fracture slip is often referred to as “shear stimulation.” When the fractures slip, their permeability is permanently enhanced, and well productivity can be improved by order of magnitude (Tester, 2007). An important challenge for the deployment of EGS is that shear stimulation triggers microseismicity, very low magnitude seismic events that sometimes can be felt at the surface (Majer et al., 2007). Induced seismicity threatens the public acceptance of EGS, and the possibility of triggering a truly damaging seismic event, while seemingly remote, deserves careful consideration. Shear stimulation directly impacts induced seismicity, the productivity of the EGS system, and the long term temperature decline, and so a practical need has developed for credible EGS stimulation modeling. We have developed a model that couples (1) fluid flow and (2) rate and state friction for the purpose of simulating EGS stimulation. This paper summarizes the use of our model to investigate induced seismicity during injection into a single, isolated fracture. The advantage of a model that couples rate and state friction and fluid flow is that it is a physically based technique that allows direct modeling of the frictional phenomena that cause induced seismicity. Rate and state friction is widely used in earthquake modeling. It has been very successful at replicating both laboratory observations of rock friction and large scale earthquake phenomenon (Dieterich, 2007, Segall, 2010). It relates friction to fracture sliding velocity and past sliding history, and provides a mechanism for both slow, aseismic slip and rapid, seismic slip. Often in EGS modeling, friction is assumed constant, which makes it impossible to model seismicity. Seismicity occurs due to high rate slip caused by rapid weakening of friction. Another requirement for modeling seismic slip is that the heterogeneous stresses induced by slip must be described accurately. Seismicity is a process where slip nucleates at a given location and rapidly propagates due to induced stresses, and so accurate description of induced stresses is critical. Some recent EGS modeling work has used a simplified treatment of friction in which sliding elements have a constant nonsliding resistance to slip, and once slip initiates an instantaneous drop in either friction (McClure and Horne, 2010) or stress (Baisch et al., 2010) is imposed. We refer to the approach in McClure and Horne (2010) as “static/dynamic friction.” Compared to static/dynamic friction, rate and state friction is a considerably more robust approach. Unlike static/dynamic it can model aseismic slip. Static/dynamic friction relies on the unrealistic assumption that all slip during a seismic event is simultaneous and instantaneous. Finally, in this paper we show how static/dynamic friction simulation can converge to an unrealistic result for an increasingly fine scale discretization. We compared the modeling described in this paper to our earlier results from McClure and Horne (2010) and found that the results were similar in most respects. Despite the potential pitfalls of static/dynamic simulation, we conclude that because of its reasonable performance and relative efficiency, it may be appropriate for some applications. Coupling between slip and fluid flow can occur for a variety of reasons. Slip can induce an increase in both fracture permeability and pore volume. Also, slip on one fracture can affect the normal traction on another, directly changing fluid pressure. Fluid flow affects slip because frictional strength depends on fluid pressure. We examined the effect of shear-induced pore dilation, injection schedule, and the characteristic displacement scale dc on the resulting seismicity. Shear-induced pore dilation tended to damp out seismic events, resulting in more, lower magnitude events. We studied a variety of factors related to injection schedule. A higher injection pressure resulted in a greater number of events. Decreasing injection pressure over time resulted in a significantly smaller maximum event magnitude than any other injection schedule. Seismicity was triggered at the periphery of the stimulated region after injection stopped due to redistribution of fluid pressure. Producing fluid from the well after injection inhibited post-injection seismicity. Our findings were consistent with Baisch et al. (2006) who predicted that post-injection seismicity should occur due to a redistribution of fluid pressure and that producing fluid after injection should inhibit post-injection seismicity. Our injection schedule results were consistent with our prior findings using static/dynamic friction in McClure and Horne (2010). The effect of increasing characteristic displacement scale, dc, was to damp out seismic slip, as predicted by rate and state theory. Our eventual goal is to perform efficient coupled rate and state and fluid flow simulation on a complex network of multiple fractures. In the appendix we provide many of the techniques we are implementing to improve computational efficiency.