Physics-Informed Reduced-Order Modeling for Two-Phase Reservoir Simulation via Operator Inference and Hybrid LSPG Stabilization

Abstract Simulating multiphase flow in petroleum reservoirs is computationally expensive. This paper presents a physics-informed framework for building efficient and stable surrogate models for two-phase flow by reformulating the governing equations into a quadratic descriptor system. This structure enables an Operator Inference (OpInf) approach to directly learn the system's physical parameters from snapshot data by solving a small linear least-squares problem. For time integration, we construc