Assimilation of Microseismic Data into Coupled Flow and Geomechanical Reservoir Models with Ensemble Kalman Filter

Abstract

Abstract Hydraulic stimulation of tight unconventional and geothermal reservoirs has been observed to trigger microearthquakes (MEQs). Triggering of MEQ events has been linked to pore pressure, temperature, and in-situ stress variations. The resulting clouds of micro-seismic events are believed to carry information about the underlying coupled flow, geomechanics, and thermal processes and, hence, rock hydraulic and mechanical properties. We develop a framework for integrating microseismic events as monitoring data to infer reservoir property distributions. To model the reservoir stimulation and induced microseismicity, we use a fully coupled thermo-poroelastic model with coupled heat transport, fluid flow, and rock deformation capabilities as forward model and apply the ensemble Kalman filter (EnKF) to assimilate MEQ measurements. Because discrete MEQ events are not amenable to continuous estimation methods, we first use a kernel density estimation (KDE) method to convert the observed cloud of MEQ events into a continuous seismicity density map. We then apply a variant of the EnKF to integrate the resulting continuous seismicity density map to estimate heterogeneous hydraulic and geomechanical rock property distributions such as permeability, Young's modulus, tensile strength and Cohesion. We demonstrate that assimilating the large correlated datasets of seismicity density map with the EnKF can result in substantial loss of ensemble spread. To mitigate this issue, we investigate three alternative EnKF implementations: 1) dampening the changes introduced during the update, 2) projecting the microseismic data onto a low-dimensional subspace that is defined by left singular vectors of the perturbed observations matrix, and 3) using coarse-scale continuous representation of the microseismic data. We use numerical experiments to evaluate the performance of these methods.