A novel framework for dynamic risk analysis of CO2 sequestration in shallow aquifers

This manuscript introduces a novel framework for dynamic risk analysis of CO2 sequestration in aquifers. The framework integrates advanced spatio-temporal simulations and probabilistic models to deliver dynamic risk profiles. The framework incorporates Least Squares Support Vector Machines (LSSVM) and Particle Swarm Optimization-Artificial Neural Networks (PSO-ANN) to predict CO2 storage capacity and behavior. The framework's application demonstrates a cumulative storage capacity of 3.72 kilo tonnes (kt) over a century in the geological aquifer. Integrating Dynamic Bayesian Networks (DBN) in the framework enables the study of CO2 leak risk profile as a function of time. The framework underscores the importance of structural and stratigraphic trapping mechanisms, achieving a peak retention rate of ∼51.43% during the injection phase and maintaining an average retention rate of ∼47.95% across the operational timeframe (100 years) for the geological aquifer. The dynamic risk profiles provide a deeper understanding of evolving storage conditions, enabling early detection of anomalies and proactive operational adjustments. The framework's application also highlights the effectiveness of adaptive pressure management in sustaining aquifer stability and optimizing CO2 retention. The proposed framework (with integrated models) provides a unique and significant advancement in the study of carbon sequestration strategies, providing actionable insights for policymakers and industry stakeholders.