A Graph–Physics-Constrained Fast State Estimation Method for Wind/PV Integrated Transmission Networks

The increasing penetration of wind and photovoltaic (PV) generation introduces frequent operating-point variations into transmission networks, while missing supervisory control and data acquisition/phasor measurement unit (SCADA/PMU) measurements and bad data may further weaken the reliability of online state estimation. Conventional weighted least squares (WLS) estimators have a clear physical interpretation, but repeated online matrix solutions may become burdensome in large-scale rolling estimation. To address this issue, this paper proposes a graph–physics-constrained fast state estimation method with bad data detection (BDD) and filtering. Wind/PV-load operating scenarios are constructed on standard test systems, and mixed SCADA/PMU measurements are represented with missing masks and bad data perturbations. The filled measurements, measurement availability mask, and residual anomaly scores are used as input features, while the network topology is converted into a graph Laplacian prior. A regularized fast mapping, graph Laplacian smoothing, and threshold-calibrated residual screening are combined to obtain online state estimates and bad data labels. Five-seed case studies compare the proposed method with WLS and Huber robust WLS. In case300, the average online time is reduced from 66.063±3.054 ms for WLS to 1.922±0.374 ms for the proposed method, corresponding to a speedup of about 35.08 times. The results indicate that the proposed linearized prototype is most promising as a fast large-scale rolling estimator or abnormality screener, rather than as a full replacement for model-based estimators in all scenarios.