Spatiotemporal shapley value-based pressure signal decomposition for enhanced geological carbon sequestration monitoring under uncertainty

Abstract

Geological Carbon Sequestration (GCS) involves capturing CO₂ from anthropogenic sources, such as power plants and industrial processes, and injecting it into geological formations for permanent storage. Monitoring subsurface CO₂ migration is essential to ensure that the injected CO₂ remains safely sequestered and does not leak into the atmosphere. Pressure sensing, in particular, is a cost-effective and efficient method for monitoring large pore networks and detecting changes in subsurface conditions. However, the presence of multiple CO₂ injector wells operating under distinct conditions, such as varying injection rates, well locations, and completion designs, complicates the pressure response observed in monitoring wells. This complexity makes it challenging to accurately track individual CO₂ plumes originating from specific injector wells. Understanding the pressure dynamics is crucial for ensuring the integrity of the storage site and optimizing injection strategies. To address this challenge, this study proposes a comprehensive workflow for bottomhole pressure (BHP) decomposition. We utilize Shapley values, combined with geostatistical modeling and numerical flow simulation, to determine the individual pressure contributions from each injector well to the monitoring wells. By discretizing Shapley values in both time and space for a given subsurface model, we calculate the marginal pressure contributions of injector wells while accounting for interaction effects, spatial context, and time-varying operational conditions. This approach enhances the accuracy and reliability of GCS monitoring. Additionally, partial dependency plots are created to evaluate the pressure dynamics between injectors and monitor BHP over time, providing valuable insights into the behavior of the storage reservoir.