Real-time Forecasting and Operational Control of Perceivable Induced Seismicity in Geo-Energies

Linus Walter , Auregan Boyet, Ioannis Stefanou, Victor Vilarrasa

Geo-energies, including enhanced geothermal systems, geological carbon storage, and underground hydrogen storage, are essential for decarbonizing the energy and heavy industry sectors. However, their widespread implementation is hindered by the risk of perceivable induced seismicity. This review begins by evaluating existing seismic hazard mitigation schemes, particularly the traffic light system (TLS) and its advanced variants. While these schemes are practical, they have proven unreliable in fault-dominated reservoirs, as demonstrated by the shutdown of the Pohang EGS following a Mw 5.4 event. Building on these lessons, we propose a paradigm shift: The next generation of induced seismic hazard management schemes has to fully integrate data processing, probabilistic physics-based reservoir models, and operational response schemes into a single workflow. To this end, we review machine learning (ML) techniques for creating surrogate models trained on physics-based simulations and discuss how linking these models to observed seismicity enables near real-time forecasting. Such hybrid workflows would empower geo-energy operators to make informed, scenario-based decisions, balancing seismic risk with economic profitability. Additionally, we examine recent advances in applying control theory to geo-energy operations, aiming to prevent induced seismicity while maximizing production. These integrated workflows hold the potential for active operational control and possibly full mitigation of perceivable induced seismicity, paving the way for large-scale, socially acceptable deployment of geo-energies.