Diagnosis Method for Partial Discharge Faults in Power Cables Based on Deep Learning

Aiming at the current problem of low accuracy of partial discharge fault diagnosis of power cables, this paper proposes a deep learning-based partial discharge fault diagnosis method for power cables. Firstly, the causes and types of partial discharges in high-voltage transmission cables are analyzed, and an experimental platform for partial discharges in high-voltage transmission cables is constructed to collect the original signals. Then the improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN) is used to decompose the signal to obtain the reconstructed signal, and the fuzzy entropy value is introduced to construct the partial discharge fault feature vector. Finally, a Lévy flight and dynamic weights are used to improve golden sine algorithm (GSA), so IGSA is proposed for realizing the hyper-parameter optimization of convolutional neural network-bi-directional gated recurrent unit (CNN-BiGRU), so as to obtain the partial discharge fault diagnosis model of power cable based on IGAS-CNN-BiGRU. The experimental results prove that the diagnosis accuracy of the method proposed in this paper reaches 98.3333% and the diagnostic efficiency is high, which realizes the accurate diagnosis of partial discharge faults in power cables, and can give full play to the engineering efficiency in power cable operation and maintenance work.