文献类型：期刊文献

英文题名：Partial Discharge Pulse Segmentation Based on Clustering by Fast Search and Find of Density Peaks [基于密度峰值聚类算法的局部放电脉冲分割]

作者：Zhu Y. Jiang W. Liu G.

通信作者：Zhu, Y.[1]

机构：[1]State Key Laboratory of Alternate Electrical Power System with Renewable Energy Sources, North China Electric Power University, Baoding, 071003, China;[2]School of Electrical and Information Engineering, Guizhou Institute of Technology, Guiyang, 550003, China

第一机构：State Key Laboratory of Alternate Electrical Power System with Renewable Energy Sources, North China Electric Power University, Baoding, 071003, China

年份：2020

卷号：35

期号：6

起止页码：1377-1386

外文期刊名：Diangong Jishu Xuebao/Transactions of China Electrotechnical Society

收录：Scopus(收录号：2-s2.0-85083035007)

语种：英文

外文关键词：Adaptive energy threshold; Clustering by fast search and find of density peaks (DPC); Otsu adaptive energy threshold; Partial discharge; Pulse segment

摘要：Partial discharge (PD) signal processing is the basis of insulation state evaluation of electric power equipment, and feature extraction is the key link of signal processing. The feature extraction includes two steps: partial discharge pulse segmentation and discharge feature extraction. Pulse segmentation extraction is the premise of subsequent feature extraction and fault classification of PD signals. In order to preserve discharge information as much as possible and reduce manual interference, a PD pulse segmentation method based on clustering algorithm is proposed in this paper. The wavelet decomposition algorithm is used for filtering, and the noise rejection ratio (NRR) is used to characterize the filtering effect. The method takes all the partial discharge half-wave pulses into consideration to calculate the energy of each half-wave which is the integral of squared instant value of the PD signal over time. Therefore, it can describe the process of partial discharge more accurately. The Otsu algorithm is used to calculate the energy threshold adaptively, and one density peak clustering algorithm, called as clustering by fast search and find of density peaks (DPC), is used to achieve automatic segmentation of PD pulses. Three types of partial discharge models are established in the laboratory. Data from 10 sets of corona discharge, 11 sets of suspension discharge and 30 sets of cone-plate discharge are collected to verify the method. The results have achieved a recognition rate of more than 80%, which is higher or equivalent to the existing similar algorithms, demonstrating the superiority of this method. ? 2020, Electrical Technology Press Co. Ltd. All right reserved.