Physics-Informed Neural Network for Inertia Estimation of Power System with Inverter-Based Distributed Generation

Inertia is vital to guarantee power system stability and to improve power grid operations, especially with the increasing penetration of inverter-based distributed generation (DG). The reconstruction of grid frequency upon contingencies can be used to analyze system stability and estimate the power system inertia for appropriate inertia control design. This paper proposed a physics-informed neural network (PINN) to reconstruct both the grid frequency and its rate of change of frequency (ROCOF) required to estimate system inertia by solving a power regulation problem formulated as a partial differential equation (PDE), whereby the residual of the PDE is added to the loss function of the physics-informed neural network during training. The proposed PINN can learn the solution to the power regulation problem. System inertia is estimated from the power perturbation data, reconstructed frequency, and ROCOF. This method is validated using the IEEE 39 bus system modelled in MATLAB/Simulink.