Incipient Fault Identification in Distribution Lines Based on Deep Convolutional Neural Networks

To address the limitations of traditional detection methods in identifying incipient faults in distribution lines, such as low accuracy and insufficient generalization capability, this study proposes an intelligent fault identification method based on deep convolutional neural networks (DCNN). The proposed method constructs a multi-level convolutional architecture, integrating optimization strategies like batch normalization, residual connections, and Dropout to effectively capture timefrequency characteristics of fault signals, thereby enhancing the model’s generalization ability and robustness. Using fault recording data from distribution lines, the study focuses on classifying three types of incipient faults: cable faults, insulator faults, and tree-contact faults. Model performance was evaluated through 4 -fold cross-validation and independent test sets. Experimental results demonstrate that the proposed method achieves an average accuracy of $97.67 \%$ on the test set, significantly outperforming traditional support vector machine (SVM) and long short-term memory (LSTM) models. Notably, it attains $100 \%$ accuracy in insulator fault identification. The method proves effective in detecting subtle fault signatures, offering a novel technical solution for precise incipient fault detection in distribution lines with high practical engineering value.