The Fault location method for distribution networks based on graph convolutional network

Fault detection and localization in distribution networks are of significant importance for ensuring the stable operation of power systems. With the continuous integration of renewable energy sources and distributed power generation, modern power systems are becoming increasingly complex, rendering traditional fault localization methods inadequate in providing accurate assurance. This paper proposes a fault location algorithm for distribution networks based on graph convolutional neural networks. By setting certain nodes in the distribution network as observation nodes, this algorithm captures voltage data features of these observation nodes using graph convolutional neural networks, thereby achieving precise fault localization across the entire distribution network. Simulation on the IEEE-33 node feeder system demonstrates that the fault localization algorithm based on graph convolutional neural networks outperforms other methods in terms of performance.