局部放电
超高频
开关设备
声学
计算机科学
工程类
电气工程
遥感
电子工程
电极
材料科学
探测器
解调
信号处理
作者
Hakima Cherif,Ciptian Weried Priananda,Hazlee Azil Illias,Jun Qiang Chan,Wong Jee Keen Raymond,Ali Chemsa
标识
DOI:10.1109/tdei.2026.3699319
摘要
Partial discharge (PD) monitoring is essential for the condition assessment of air-insulated switchgear (AIS). This paper presents an experimental framework that acquires ultra-high-frequency (UHF) PD signals corresponding to four representative insulation defects: corona, surface, floating electrode, and triple junction. The collected signals are transformed into time–frequency representations using the Continuous Wavelet Transform (CWT) and the Wavelet Synchrosqueezing Transform (WSST) with two analytic mother wavelets, namely Amor and Bump. The resulting scalograms are employed as inputs to a convolutional neural network (CNN) for training and are evaluated using 5-fold cross-validation for PD classification. Comparative results show that WSST consistently improves class separability over CWT across both wavelets, where the WSST-Amor configuration achieves the best overall accuracy of 98.02%, surpassing CWT-Amor, WSST-Bump, and CWT-Bump. Compared to traditional phase-resolved partial discharge (PRPD) approaches, which require larger datasets and more complex acquisition setups, the proposed scalogram-based method provides rich feature representations using fewer signals, offering a more practical and efficient pathway for automated PD diagnosis in AIS.
科研通智能强力驱动
Strongly Powered by AbleSci AI