A Novel Distribution Network Fault Location Method Based on Improved Convolutional Neural Network

Addressing the complexities and accuracy demands of fault localization in current distribution networks, accurate and timely fault diagnosis is crucial for the safe operation and reliability of power supply systems. However, traditional deep learning methods have limited the exploration and utilization of electrical network data characteristics. Therefore, this paper proposes a novel fault localization method for distribution networks based on an improved convolutional neural network (CNN). This method includes several innovations and improvements: Firstly, considering the unique structure and fault characteristics of distribution networks, a specific three-dimensional CNN structure was designed. This network can process and analyze visualized three-dimensional datasets from distribution networks, enabling more effective extraction of crucial features and information from the image domain. Moreover, this improvement is built on the foundation of CNN, resulting in a smaller network structure that occupies less memory compared to other architectures.