Machine Learning-Based Prediction of CO₂ Storage Capacity and Injectivity

Abstract

Accurate prediction of CO₂ storage capacity and injectivity is fundamental to the technical and economic viability of geological carbon storage (GCS), governing site screening, well design, and long-term project planning across depleted hydrocarbon reservoirs, deep saline aquifers, and geothermal-analogue formations. Conventional approaches to capacity and injectivity estimation rely on physics-based reservoir simulation and empirical volumetric methods, both of which are constrained by high computational cost, extensive data requirements, and limited scalability across the large number of geologic realizations needed for robust uncertainty assessment. Machine learning (ML) has emerged as a powerful complementary approach, capable of learning nonlinear relationships between petrophysical, geomechanical, and operational parameters and their resulting storage capacity or injectivity outcomes directly from simulation or field data, often at a fraction of the computational cost of conventional methods. This review synthesizes recent advances in ML-based prediction of CO₂ storage capacity and injectivity, examining artificial neural network, ensemble tree-based, and hybrid architectures applied to storage efficiency estimation, trapping index prediction, and permeability-based injectivity assessment. It draws on case studies from heterogeneous depleted gas reservoirs, geothermal-analogue petrophysical characterization, and geomechanically informed wellbore stability assessment, and further considers methodological parallels with large-scale AI-driven predictive analytics and economic impact frameworks developed for national healthcare systems. The review concludes by identifying persistent challenges in training data scarcity, cross-site generalizability, and integration of geomechanical constraints, and proposes a research agenda for advancing ML-based capacity and injectivity prediction toward routine use in commercial-scale CO₂ storage site screening and design.