Research on intelligent inspection method for power grid anomalies based on state estimation and deep dual-path network

With the rapid development of new power systems, cybersecurity threats have become increasingly prominent, particularly false data injection attacks (FDIA), which pose severe challenges to the stable operation of power systems. To address the limitations of existing anomaly detection methods in dynamic system state estimation and classification accuracy, this paper integrates weighted least squares (WLS) and extended Kalman filtering (EKF) to achieve high-precision state estimation, while employing χ²-test and anomaly detection indices to identify abnormal behaviors. For fine-grained classification of anomalous events, a deep neural network is designed by combining Convolutional Neural Network (CNN) and Bidirectional Long Short-Term Memory network (BiLSTM), where CNN extracts spatial features and BiLSTM captures bidirectional temporal dependencies, thereby improving model accuracy. Experimental validation on the IEEE 14-bus system demonstrates that the proposed method significantly outperforms traditional CNN approaches in both single-node and multi-node attack scenarios in terms of accuracy and F1-score.