Advancing Surrogate Modeling for CO2 Storage Risk Through Cutting-Edge Survival Analysis-Driven AI
International Petroleum Technology Conference (IPTC) Summit on AI for the Energy Industry, 2026
We advance survival analysis-based surrogate modeling for geological CO2 storage risk using modern machine learning methods, including optimal survival trees. By training on shortened reservoir simulations and accounting for censored observations, these models predict time to leakage while reducing the computational cost of generating training data.
We evaluate the framework through a synthetic saline aquifer study and a reservoir simulation based on the SACROC site in West Texas. These applications examine the accuracy and interpretability of survival-based surrogates under geological uncertainty and realistic operating conditions, supporting decisions about injection control, pressure management, and long-term storage risk.
Recommended citation: Gurwicz, A., Gutman, D. H., & Gildin, E. (2026). Advancing surrogate modeling for CO2 storage risk through cutting-edge survival analysis-driven AI. IPTC Summit on AI for the Energy Industry, Dubai, United Arab Emirates, January 13–14. Paper IPTC-25222-MS. https://doi.org/10.2523/IPTC-25222-MS