Machine Learning Application to Underground CO2 Storage

This study develops a machine learning framework to predict the strength degradation of sandstone under CO₂ exposure. By training separate models for UCS and BTS change rates, and coupling them to simulate time-dependent behavior, a predictive envelope for CO₂ injection safety is constructed. The results highlight a narrowing strength window over time, providing a quantitative basis for defining safe injection pressure limits. This coupled approach bridges laboratory data with practical applications in wellbore stability evaluation for CCS projects.