Fault Location Method of Active Distribution Networks Based on Graph Attention V2 Network

This paper develops a fault location method for active distribution networks fault location method based on the graph attention v2 network (GATV2). The GATV2 introduces the concept of dynamic attention and improves the performance of the model by adjusting the order of operation of the network layer of the graph attention network (GAT) to achieve the calculation of the dynamic attention value. The efficacy of the GATV2-based fault location model of active distribution networks is validated through the construction of a fault scenario model for active distribution networks within the IEEE33 node network. The experimental outcomes demonstrate that the positioning accuracy and generalization performance of the GATV2 model is superior to that of the traditional fault location model when performing traditional single-phase ground fault scenarios and fault scenarios where multiple fault types coexist.