A Fault Diagnosis Algorithm for Distribution Networks Based on Graph Convolutional Neural Networks

As power grids continue to expand, their structures become more complex, leading to an increased frequency of multiple faults, which presents a critical challenge for fault diagnosis. To tackle the issues of handling large volumes of fault data and the need for rapid and accurate fault diagnosis in power grids, this paper introduces a fault diagnosis model for distribution networks based on a fuzzy-optimized graph convolutional neural network (GCN). The methodology involves several steps: Data Processing: Collected characteristic data from distribution network fault lines are preprocessed. Model Construction: A fault diagnosis model based on GCN is developed, and the membership function for distribution network faults is established using fuzzy theory. Model Application: The trained model is utilized to diagnose distribution network faults. Simulation outcomes demonstrate that the fuzzy-optimized GCN achieves higher accuracy in multi-fault diagnosis compared to conventional neural networks and alternative methods. The proposed method yields more precise diagnosis results, and overall, its diagnostic performance is superior..