Multiresolution Fourier Neural Operator Framework: A Novel Approach for Hydro-Geochemical Coupling Modeling in CO2 Geological Storage

Summary CO2 Geological Storage (CGS) serves as a critical decarbonization technology globally. The injection of acidic CO2 disrupts water-rock equilibrium, initiating geochemical reactions that alter reservoir porosity and permeability. Accurate simulation of pressure evolution and CO2 plume migration demands multiresolution modeling, yet conventional numerical approaches suffer from prohibitive computational costs. This study addresses this challenge through a novel multiresolution Fourier neural operator (MRFNO) framework. The framework enables concurrent prediction of hydrodynamic and geochemical field dynamics, efficiently characterizing near-well permeability evolution driven by CO2-water-rock interactions across spatial scales. The methodology develops FNO-based accelerated solvers for governing partial differential equations in both hydrodynamics and geochemistry. Field coupling is implemented via inter-model parameter transfer at synchronized timesteps, with porosity-permeability updates from geochemical reactions computed using the Verma-Pruess model. MRFNO architectures are trained under physics-informed constraints that enforce multiscale consistency and cross-field coupling mechanisms within the loss function. This integration establishes an intelligent simulation system for coupled hydro-geochemical processes at adaptive resolutions. Validation on a 200-year field-scale 3D scenario (20-year injection + 180-year post-injection monitoring) demonstrates breakthrough performance: The framework achieves nearly 2000-fold acceleration compared to traditional methods while maintaining prediction errors below 3% mean absolute percentage error for CO2 saturation, brine density, pressure, and mineral volume fields. Near-well permeability and CO2 saturation predictions exhibit determination coefficients (R2) exceeding 0.98 across multiple resolutions. This innovation provides transformative capabilities for CGS projects, delivering high-fidelity simulation of coupled hydro-geochemical processes and quantitative characterization of formation damage induced by water-rock reactions. The framework establishes a new paradigm for real-time, precision modeling to support industrial-scale CGS deployment.