Single-phase Line Breakage Fault Location for Active Distribution Systems Based on Inductive Graph Neural Network

In order to improve the situation awareness of distribution systems under extreme weather, this paper studied the single-phase line breakage fault location for active distribution systems based on the inductive graph neural network. In this paper, the graph sampling aggregation (GraphSAGE) algorithm is applied to accurately locate the line breakage fault of the distribution systems, which can fully mine graph features, predict unknown nodes based on known node information. Firstly, the effect value of the three-phase current on each line after the fault in a distribution system is obtained as the node feature in datasets. Then the GraphSAGE model is built and the training set is used to train the model. Finally, the feasibility and accuracy of this model is validated by a study on a modified IEEE 123-bus system. Compared with graph convolutional neural networks (GCN) and graph attention networks (GAT), it is verified that GraphSAGE is more universal and can better adapt to the changes of system topology.