Enhancing CO2 Storage Modeling in Subsurface Energy Systems with Nested Fourier

Abstract This paper explores the application of Nested Fourier Neural Operators (FNOs) to enhance the accuracy and computational efficiency of CO2 storage modeling in subsurface energy systems, focusing on their ability to model complex fluid flow and geomechanical interactions. The study employed a hybrid approach that integrates traditional numerical methods with Nested Fourier Neural Operators to model the dynamic processes involved in CO2 storage. The FNO The framework was trained on extensive datasets from field experiments and high-fidelity simulations, learning complex relationships between fluid flow, reservoir properties, and geomechanical responses. Comparative analyses with conventional numerical solvers were performed to evaluate model accuracy and computational speed improvements. Implementing Nested Fourier Neural Operators significantly reduced computational time, achieving a 40% decrease compared to conventional solvers while accurately predicting CO2 plume migration and storage behavior. The FNO approach also enhanced the model's ability to capture non-linear interactions between reservoir parameters, leading to more precise forecasting of long-term storage stability. Sensitivity Analyses revealed the model's robustness across geological conditions and injection scenarios. This study introduces the application of Nested Fourier Neural Operators to CO2 storage modeling, offering a novel and efficient approach to improving the accuracy and scalability of subsurface energy system simulations.