Assessment and optimization of maximum magnitude forecasting models for induced seismicity in enhanced geothermal systems: The Gonghe EGS project in Qinghai, China

Abstract

Seismic activity induced during the development of Enhanced Geothermal Systems (EGS) is frequent and poses significant hazards. This study aims to accurately forecast the maximum magnitude (M_max) of induced earthquakes to effectively manage seismic risks. Focusing on the EGS project in Gonghe County, Qinghai Province, we evaluated and optimized various widely-applied M_max forecasting models, while also endeavoring to directly forecast maximum magnitudes in the post-closure phase. Initially, advanced deep learning models (such as PhaseNet and GaMMA) were employed to process seismic data, coupled with VELEST and HypoDD methods for earthquake relocation. Subsequently, four currently widely recognized maximum magnitude (M_max) forecasting models (H14, NRBE, V16, and G17) were utilized to forecast and assess M_max during nine hydraulic fracturing stages, six post-closure stages exhibiting tailing effects, and the entirety of the 2019–2021 period in the Gonghe EGS project. The findings indicate significant disparities in the efficacy of different forecasting models during hydraulic fracturing stages, with no model fully aligning with the complex physical mechanisms of induced seismicity. NRBE and G17 models tend to overestimate M_max forecasting, potentially escalating production costs, whereas H14 and V16 models yield results closer to actual values but are susceptible to the influence of real seismic breakthroughs. Furthermore, distinct discrepancies were observed in the M_max forecasting performance of the same model between hydraulic fracturing and post-closure stages. Attempts to directly forecast M_max post-closure achieved certain efficacy, likely due to the cumulative injection volume exerting a degree of control over induced seismic activity in both stages. Lastly, to overcome limitations in current M_max forecasting models, a hybrid model Y24, integrating the advantages of four forecasting models, was proposed, demonstrating higher accuracy and reliability in forecasting during both hydraulic fracturing and post-closure stages. The study's findings provide crucial technical support and decision-making basis for the seismic risk management of EGS projects or shale gas development projects employing hydraulic fracturing, underscoring their significance in ensuring the safety and sustainability of new energy and resource development endeavors.