Fault Section Location in Active Distribution Networks Based on Traveling Wave and Graph Convolutional Network

The large-scale integration of distributed power sources has transformed the power flow structure of traditional distribution networks from unidirectional to multidirectional, increasing system complexity and posing challenges for fault location. Traditional methods struggle to handle the complex network topology and bidirectional power flow structure introduced by multiple power sources, resulting in reduced accuracy and robustness in fault location. Although deep learning techniques hold great potential for fault location, existing studies primarily focus on analyzing the features of individual nodes, neglecting the importance of the complex topological information in distribution networks, which is crucial for improving location accuracy. To address this, this paper proposes a fault location method based on Graph Convolutional Networks (GCN) and voltage traveling wave signals. Voltage traveling wave signals can rapidly transmit fault information when a fault occurs, while GCN can perform deep feature extraction considering the topological structure of the distribution network. By integrating voltage traveling wave signals and GCN, this method achieves precise fault section location in active distribution networks, and experimental results validate its effectiveness in such networks.