Accelerating Power System Dynamic Simulations with Physics-Informed Neural Networks

Physics-informed neural networks (PINNs) have emerged as powerful tools for solving complex power system problems by integrating physical models. This paper applies PINNs to solve differential equations governing the Northeast Power Coordinating Council (NPCC) system. The approach adaptively transitions between the original nonlinear model, a trained PINN, and a hybrid model, partitioning the system into a detailed study area and a simplified external area. It uses a fully nonlinear model during the fault-on period, a hybrid model for large rotor angle deviations, and PINNs for smaller state variations. Simulation results are compared with cases using the original nonlinear model and an adaptive switching strategy between hybrid and linear models. PINNs enable faster, more accurate simulation results than conventional methods, efficiently computing dynamic states with less reliance on extensive training data.