Development of 1D Hybrid Hydromechanical Models for Real-Time Forecasting of Induced Seismicity Rate

Abstract

Hydraulic stimulations play an important role in Enhanced Geothermal Systems (EGS) by increasing the permeability of the host rock and facilitating more efficient fluid circulation and heat extraction. However, fluid injection operations are unavoidably accompanied by induced earthquakes. Adaptive Traffic Light Systems (ATLS) have been proposed as seismic risk mitigation tools for EGS stimulations. An ATLS scheme aims to provide real-time, adaptive, and time-dependent probabilistic seismic forecasts by leveraging the latest available data during ongoing industrial operations. Critical to ATLS are numerical models capable of robustly forecasting the temporal evolution of induced seismicity, while properly accounting for uncertainties. In this work, we present two classes of 1D hybrid hydromechanical models for real-time forecasting of induced earthquakes. We retrospectively apply these models to data sets from hydraulic stimulations performed at four different spatial scales: Grimsel Test Site (2017), Bedretto Underground Laboratory for Geoenergy and Geosciences (2022), Utah FORGE (2022), and Basel Deep Heat Mining project (2006). We compare the models' forecasting performance and real-time applicability. We found that a nonlinear pressure solution that accounts for both reversible and irreversible permeability changes, coupled with an analytical probability density-based approach to simulate seismicity, is more suitable for industrial-scale applications. A stochastic approach that explicitly simulates seismicity, albeit simplified and improved for computational efficiency, exhibits greater variability in performance and remains computationally expensive for industrial-scale cases involving large seismic data sets and high spatial resolution requirements.