Fault Detection in Power Communication Networks Based on GCN-GRU

With the increasing scale and complexity of power communication networks, traditional fault diagnosis methods are inadequate for dealing with dynamic, multi-dimensional, and massive data. To address these challenges, this paper proposes a dynamic fault detection model based on a graph convolutional network (GCN) and gated recurrent unit (GRU). By modeling devices in the power communication network as graph structures and combining GCN to extract spatial features of the network topology with GRU to capture dynamic characteristics in the time dimension, accurate fault diagnosis can be achieved. First, event codes and association matrices are constructed by encoding the fault types of power communication devices, allowing the visualization of connections between devices and fault patterns. Then, GCN is used for frequency and spectral domain feature extraction, combined with time series modeling using GRU to capture the spatial and temporal correlation of faults. The experimental results demonstrate that the proposed method outperforms traditional methods in terms of accuracy and efficiency, providing a new approach for the intelligent maintenance of power communication networks.