A Method for Recognizing and Classifying Arc Sound Signals in Low-Current Faults

Abstract

Accurately identifying faulty electrical signals in distribution networks remains challenging. However, arc sound signals provide effective features for fault recognition. Due to the complex operational environment of distribution network lines, which are susceptible to interference from environmental noise, several issues arise, including the difficulty in determining optimal parameters for noise reduction, the challenge of extracting effective features from arc sound signals, and the relatively low classification accuracy. To address these challenges, a recognition method based on architectural optimizations of the VGG16 network, combined with machine learning techniques, is proposed for arc fault recognition in distribution networks. First, the Whale Optimization Algorithm is used to optimize the parameters of Variational Mode Decomposition to denoise five types of arc sound signals collected under laboratory conditions. Next, the sound signals in the dataset are transformed into Mel spectrograms, which are then fed into the improved VGG16 neural network for feature extraction. Finally, a combined classifier is used to categorize the sound signals, facilitating fault diagnosis. The experimental results show that the proposed method significantly improves the extraction effect of arc glow sound signal features and has higher accuracy in the recognition and classification of arc glow sound signals compared to existing methods. © 2026 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.