DESC0024778

Geologic Reservoir Engineering (GRE) activities such as energy production (including oil/gas and geothermal), energy storage (including hydrogen), and waste storage (e.g., CO2 and radioactive such as the WIPP site) are critical for transitioning to a zero-carbon economy and addressing energy and environmental problems that our nation is facing. However, induced seismicity has the potential to threaten reservoir integrity, causing infrastructure damages and creating government, regulatory, and public objections that may terminate GRE projects. Existing approaches to seismic hazard characterization fail to account for subsurface complexities and thus produce unreliable seismic hazard forecasts. We aim to develop data-driven, AI-based methods and software (AI-PREDICT) to determine if future seismic events in a reservoir are imminent via novel approaches we have developed and will develop. Unlike other methods, ML approaches allow us to incorporate existing knowledge about fault physics while also ôlearningö about unknown or difficult-to-measure features. ML methods have been successfully applied to detect faults and probe their current and future slip characteristics in prototype studies. We intend to further develop ML approaches for fault detection and seismic hazard characterization by analyzing seismic data and incorporating equations representing governing seismic processes in the ML process. The Phase I work will be centered around an in situ injection experiment, located in Southern France, where existing faults are driven to failure by fluid injection. The system is densely instrumented, with fault displacement measurements from seismometers and strain meters and monitoring of fluid injection pressure. We will apply ML to the continuous seismic signal and other data sources to predict future displacement and earthquake timing and assess the overall seismic hazard. In Phase II and beyond, AI-PREDICT will (1) assess site seismic characteristics, (2) evaluate the induced seismic risks based on available geological, geophysical, and seismic data, and (3) facilitate and optimize site data acquisition for improved site seismic hazard characterization. AI-PREDICT will also (4) identify subsurface activities causing induced earthquakes, (5) pinpoint data attributes and datasets informative for seismic risk evaluation, and (6) propose mitigation activities such as pressure management to reduce the induced seismicity risks. AI-PREDICT will increase the viability and safety of GRE activities while building public trust through rigorous seismic monitoring and assessment.