Data-driven Bad Data Detection and Identification in Power System State Estimation Considering Partitioning

The existence of multiple interacting and consistent bad data is a major obstacle to the application of power system state estimation. In this paper, we propose a multi-label classifier based on a spatio-temporal graph convolutional network (STGCN) to detect and identify bad data in power systems, including gross errors, topology errors, and parameter errors. To enhance the efficiency and accuracy of detection and identification, we propose to divide the entire system into multiple subareas and applying the classifier to each subarea. Extensive simulation results demonstrate that the proposed method outperforms other machine learning and model-driven methods, offering superior accuracy, robustness, and reliability in detecting and identifying bad data.