A Real-Time Single-Phase Ground Fault Localization Method based on Improved Attention Mechanism Graph Neural Network for Multi-Source Data Fusion

In this paper, a real-time single-phase ground fault localization method for multi-source data fusion based on an improved attention mechanism graph neural network (AM-GNN) is proposed. The method introduces an adaptive attention mechanism based on graph neural network (GNN), which enhances the ability to recognize key nodes and edges. Meanwhile, the accuracy and real-time fault localization are improved by integrating multi-source data and dynamically adjusting the importance of different data sources by using the attention weights. In addition, this paper adopts the spatio-temporal attention mechanism (STAM) to effectively handle multi-source temporal data, utilizing temporal convolutional networks and multi-head self-attention mechanism to capture the spatio-temporal characteristics of the data. The deep integration of AM-GNN and STAM is realized through an innovative two-way information flow and attention-guided feature fusion mechanism. Compared with traditional GNN methods, the method proposed in this paper better captures complex interactions in the system and demonstrates greater scalability when dealing with large-scale power grids. Experimental results show that the method outperforms existing methods in terms of fault localization accuracy and response time. This study provides a new solution for real-time monitoring and localization of single-phase ground faults in power systems, which is of great significance for improving the safety of power grid operation.