Fault Location Technology of DC Control and Protection System Based on Deep Learning

Power systems sometimes experience various types of faults, and fault location in power systems has become increasingly complex with the growing complexity of distribution networks and the diversity of measurement data. This study presents a fault location technique for power systems utilizing a Graph Convolutional Neural Network (GCN), enhanced with an attention mechanism to boost both accuracy and stability in fault detection. By collecting real power grid data, fault data is simulated for model training and testing. Measurement points are treated as nodes in the graph, and node connections are constructed based on the power grid structure. The GNN is used to interact node information, while a Transformer model with an attention mechanism aggregates and predicts the information. Additionally, the paper compares the proposed model with several existing methods, demonstrating the accuracy and stability of the model for fault location in power systems.