Anomaly Detection of Cyber Attacks in Smart Grid Communications Based on Residual Recurrent Neural Networks

The smart grid consists of widely distributed devices, and the complex, multitiered communication network structure increases its vulnerability to attacks. Attackers can exploit communication vulnerabilities between layers to launch network attacks. To optimize communication among smart grid devices and improve the accuracy of detecting network anomalies, this article proposes a deep learning-based method for detecting communication network anomalies in power grid devices. Data from standard sources, such as smart meters, are collected, and the modified flow direction algorithm (MFDA) is employed for optimal weighted feature selection. The selected features are then input into an adaptive residual recurrent neural network (ARRNN) with dilated gated recurrent units (DGRUs) for anomaly detection, improving the accuracy of abnormal traffic monitoring in the smart grid. Simulation results show significant improvements in key performance indicators, with the false discovery rate (FDR) reduced to 0.060, the Matthews correlation coefficient (MCC) reaching 0.881, and specificity, recall, and precision all at 0.940. Large-scale data experiments demonstrate outstanding performance in terms of memory usage, computational speed, and latency. This method proves to be effective and robust for communication anomaly detection in power grid networks.