Abstract

Carbon Capture and Storage (CCS) is currently the most practical and scalable solution for significantly reducing atmospheric CO₂ levels, helping to mitigate environmental risks and advance global efforts toward achieving net-zero emissions by 2050. However, CCS operations introduce geomechanical challenges such as increased pore pressure, which can lead to caprock failure, fault reactivation, poro-elastic deformation, and compromised well integrity— potentially resulting in CO₂ leakage, induced seismicity, and ground uplift. These risks could undermine public confidence in CCS initiatives. This study addresses these challenges by leveraging machine learning (ML)-based geomechanical and fault-slip analyses in a mature hydrocarbon field that has been in production since the late 1960s. The deeper formations, composed of flat epeiric carbonate ramps with mixed carbonate-evaporitic sequences, exhibit low compaction rates (axial strain: 0.0005– 0.002), making them ideal candidates for CCS. Using data from 58 wells, the ML-based workflow automates data preparation, outlier handling, and log prediction, significantly reducing manual effort and uncertainty. The resulting 1D Mechanical Earth Models (MEMs) and fault-slip assessments support robust measurement, monitoring, and verification (MMV) planning, enabling safer CCS site design through improved efficiency and reduced reliance on human expertise