Research and Implementation of Intelligent Recovery Technology for Fault Routing in Power Communication Networks Based on Deep Learning

To address the shortcomings of rapid fault detection, evaluation, and recovery in power communication networks, a deep learning-based fault routing intelligent recovery technology for power networks is developed and applied, including three steps of training data generation, neural networks offline training, and neural networks online inference. A communication-grid interdependent network cascade fault model fused with communication network modeling and dynamic power transmission network modeling is established to evaluate the risk degree of internal components of power communication networks more comprehensively and reliably. An outlier screening algorithm based on clustering and standard deviation is also proposed, which can effectively screen out the abnormal data to reduce the model calculation cost and improve the accuracy of research and evaluation. This study achieves the rapid generation of better routing recovery decisions and effectively improves the risk resistance of communication networks in power systems.