This paper presents a coupled 3D fluid-flow and geomechanics simulator developed to model induced seismicity resulting from wastewater injection. The simulator modeled several cases of induced earthquakes with the hope of providing a better understanding of such earthquakes and their dominant causal factors, along with primary mitigation controls. Implementation of rate-and-state friction to model friction weakening and strengthening during fault slip to accurately model earthquake occurrence, and an embedded discrete fracture model to efficiently model fluid flow inside the fault, are among the essential features of the simulator. The complete paper presents results from a combined model that brings together injection physics, reservoir dynamics, and fault physics to explain better the primary controls on induced seismicity.
Introduction
Since 2009, a substantial increase in the number of earthquakes in the central and eastern United States has occurred. Oklahoma has been one of the most affected regions, with several earthquakes of M 5+, including the Prague earthquake in 2011 and the Pawnee earthquake in 2016. This has prompted efforts to find and understand any correlation between oil and gas activity—mainly wastewater disposal—and the occurrence of the earthquakes. Induced or triggered seismicity associated with wastewater injection, mining, oil and gas extraction, and geothermal operations has been identified since the earthquakes at the Rocky Mountain Arsenal in 1960s. Fluid injection into subsurface formations can increase pore pressure, reduce the effective stress, and induce slip on faults. Laboratory studies show that the sliding displacement may enhance fracture transmissivity and create a hydraulic pathway through the formation. In an unconventional reservoir, the enhancement could lead to improved hydrocarbon production by using the slipped natural fractures. But in some formations, such as water-disposal aquifers, seismicity might be induced when faults are activated within the igneous basement.
Seismicity induced by fluid injection is controlled by several groups of parameters (injection, reservoir, and frictional). A fundamental understanding of which factors are the most important in triggering slip in areas of active wastewater injection and disposal has been hampered by interrelationships between the various parameters, leading to suggestions of injection volume, rate, or pressure being the most important. However, necessary reservoir characteristics, such as size and permeability, are not well characterized at the well or in the subsurface, and remain the main challenge for deterministic models. Additionally, rupture nucleation on faults near a region of injection depends on rate-and-state and related physics.
The literature contains many relevant seismicity studies, which are identified in the reference section of the complete paper. A 2D simulation of fluid flow inside the fault zone suggested that post-shut-in earthquakes are likely to occur nucleating at the fault edges. Using a similar numerical scheme, other researchers investigated the 2011 Prague earthquake sequence to model the delayed triggering mechanism between the M 4.8 foreshock and M 5.6 main shock. The result of the study contributed to defining constraints on values of fault transmissivity, fault compliance, and rate-and-state frictional properties. Although expensive, these types of full-physics models are capable of characterizing reservoir and fault properties. A 3D simulation in 2015 modeled fault activation under direct injection into the fault during shale-gas hydraulic fracturing. It found that for brittle faults, the moment magnitude can be higher.
A statistical method was proposed in 2017 to identify induced seismicity from large data sets by associating seismicity and wastewater injection along time. However, in depleted reservoirs, because seismicity usually occurs with a long delay after injection starts (depending on the stress state), application of these statistics will be challenging.
For the present study, a coupled 3D fluid-flow and geomechanics simulator was developed to model induced earthquakes. The model incorporates reservoir properties including vertical and horizontal extent; stratification including top-seal, reservoir, and basement; multiple permeability; and porosity. Injection parameters include rate and pressure. Fault properties include size, 2D permeability, and frictional properties. Several suites of simulations were run to evaluate the relative importance of each of the factors from all three parameter groups. The simulator assumes isothermal conditions which, except for modeling induced seismicity in geothermal reservoirs with high temperature gradient, pose no problem to solution accuracy. The complete paper describes the methodology used in the coupled simulator and validation of the developed tool, provides results and discussion for the simulations performed, and presents concluding remarks.
Methodology
The authors took a deterministic approach and built a physics-based simulator that includes the effect of fluid flow, displacement and stress on the fault, and dynamic friction to model unstable rupture events. The simulator consists of two main components, a 3D reservoir simulator and a 3D geomechanics simulator. The fluid-flow simulator solves for pressure diffusion inside the reservoir and fault. The main variables are pore pressure and saturation for each cell. The geomechanics simulator solves for stress and displacement on the fault. The main variables are sliding and opening displacements, and direction of sliding on each element. The rate-and-state constitutive equation is used to compute dynamic friction as fault slips. Module coupling facilitates the communication between reservoir cells and geomechanic elements to update dynamic permeability and friction coefficient. The coupling is sequential, meaning that initially fluid-flow parameters are calculated while geomechanical unknowns are held constant. Then, geomechanical unknowns are solved while pressure is constant. This provides an accurate solution and stable scheme. Finally, the energy released because of the sliding displacement is calculated to determine moment magnitude of an induced seismic event. The authors’ ultimate goal is to use this work flow to design a better wastewater injection plan that mitigates risk of an induced earthquake by providing a deterministic limit on earthquake magnitudes. The paper discusses the methodology and its validation in detail.
Results and Discussion
Coupled geomechanics and fluid-flow simulations were run to understand the effect of injection parameters, reservoir state of stress, and fault properties. The geometry used in these simulations is shown in the complete paper. Four sets of simulations were performed. The first set of simulations examined the effect of injection rate. The second set studied the effect of reservoir initial pressure. The third set considered the effect of fault permeability relative to the formation. The fourth set focused on fault distance to the injection well.
The higher injection rate released more than 10 times the initial energy compared with the lower injection rate and thus induced a larger first earthquake. Although pressure is at the critical level, which means that any pressure increase will nucleate a rupture event, a higher injection rate perturbs a larger area of the fault, creating a larger earthquake. This suggests there may be a critical injection rate for any specific reservoir-fault-injection combination which, if exceeded, will induce a large earthquake.
The simulation case with lower initial pressure took longer to induce an earthquake event because it required a higher pressure increase to compensate for initial pressure. Fault permeability controls how quickly pressure diffuses inside the fault. A higher ratio of fault to matrix permeability led to a larger initial earthquake, as expected. Higher permeability ratio means a larger area of the fault is perturbed and, therefore, a larger earthquake event would occur.
For the final set of simulations, two cases were run to gain a better insight into the effect of fault distance. In the first case, the fault was 10 m away from the injector, as if directly injecting into the fault. In the second case, the fault was 600 m away from the injector (Fig. 1). The first case released a huge amount of energy in the first hours after the start of injection. That was also the case because the fault had critical pore pressure, and any pressure disturbance triggers a slip event. The second case required several days for the pressure front to reach the fault, delaying earthquake occurrence. Additionally, because a smaller portion of fault was perturbed, the initial earthquake was smaller compared with that in the first case.
Conclusions
- Injection rate is the main driving factor of seismicity, especially in the case of constant-pressure-boundary condition, while volume is unimportant.
- However, for reservoirs with no flow-boundary condition, although injection rate is still an important factor, volume controls the average pressure inside the reservoirs and plays a role.
- Fault permeability, reservoir initial pressure, and fault distance affect how large of an area of the fault is perturbed by pressure increase, and are among key factors explaining induced seismicity. This implies that safer water-disposal operations require prioritizing those reservoirs in which fault surfaces exposed to fluid pressure are minimized. For reservoirs with subcritical stress state, this means high-permeability reservoirs are preferred, because it takes longer to exceed critical pressure. However, for reservoirs at critical stress state, low-permeability reservoirs are preferred because they hinder the pressure front from reaching the fault surface.
This article, written by JPT Technology Editor Judy Feder, contains highlights of paper SPE 191670, “Wastewater Injection and Slip Triggering: Results From a 3D Coupled Reservoir/Rate-and-State Model,” by Mohsen Babazadeh, SPE, and Jon Olson, SPE, The University of Texas at Austin, prepared for the 2018 SPE Annual Technical Conference and Exhibition, Dallas, 24–26 September. The paper has not been peer reviewed.