Online inertia estimation for wind-integrated power systems with ambient PMU measurements using sliced parallel physics-informed neural networks

The increasing integration of renewable energy sources (RES) helps reduce carbon emissions and mitigate climate change. However, the variability and low inertia of RES introduce significant challenges to frequency stability. This paper presents a sliced parallel physics-informed neural network (sp-PINN) method for online inertia estimation in wind-integrated power systems, emphasizing the application of artificial intelligence (AI) in practical electrical systems. The proposed method estimates system inertia by solving an inverse problem based on the swing equation within a PINN framework, which requires no pre-training and adapts to varying operating conditions. To address stochastic ambient disturbances, measurement data are split into multiple slices, processed in parallel, and then aggregated using a weighted average that filters out outliers based on their training loss. Case studies on the New England 10-machine 39-bus system and the Western Electricity Coordinating Council (WECC) 179-bus system demonstrate that the proposed method can accurately estimate both inertia and damping coefficients in real time, maintaining robustness under varying levels of RES penetration and measurement noise.