Coupled Fluid Flow and Geomechanical Modeling for Unconventional Field Development Optimization and Induced Seismicity Assessment

Fluid injection and extraction activities in subsurface result in disturbance of pressure and stress field underground. Such disturbance might intentionally or unintentionally cause the failure of geo-material. For example, hydraulic fracturing involves injection of large fluid volume with high injection pressure to enhance permeability and contact surface area for economically exploiting hydrocarbon from tight formation. Understanding hydraulic fracture geometry and underlying mechanisms and their impacts on well performance are key factors in the success of unconventional field development. Disposal of large volume of produced water into subsurface alters stress and pressure fields as well and it can cause the failure of faults in deep crystalline basement. The failure and slip along fault planes sometimes produce earthquakes which can damage properties on surface. Therefore, it is essential to assess and manage risks associated with fluid injection and production to minimize potential seismicity. To address the aforementioned challenges in unconventional field development and fluid-induced seismicity, this dissertation discusses studies related to coupled fluid flow and geomechanical modeling workflows. First, we developed a novel hybrid Fast Marching Method-based simulation (FMM-Sim) workflow for history matching and completion optimization of hydraulically fractured wells. We introduced pressure-dependent fracture property curves, based on empirical relationship, lab experiments and theoretical background, to mimic fracture propagation and closure in reservoir simulations. Therefore, we can capture the impact of completion design such as injection fluid volume and cluster spacing on the well performance. Second, we built a hybrid model, combining physics-based reservoir simulations and machine learning algorithms for unconventional reservoirs. The simulation input and output were incorporated into machine learning algorithms such that the algorithms can learn underlying physics between simulation input (e.g. completion design) and output (e.g. cumulative oil production). The hybrid model provides fast and scalable applications for unconventional field development with high accuracy which makes the hybrid modeling approach more suitable for field applications. Third, we utilized one-way coupled fluid flow and geomechanical simulations with detailed fault modeling in the Azle area, North Texas to quantitatively assess potential for induced seismicity. We incorporated fault geometry based on the operator’s seismic survey into flow and geomechanical simulations and assessed fault slip and fault frictional energy. The fault frictional energy was then compared with the radiated energy from the observed earthquakes to build empirical seismological model. The seismological model was used to predict earthquake frequency with different injection/production operations