基于能量谱分析和DBSCAN改进算法的多源局部放电分类识别方法  

Multi-source Partial Discharge Classification and Recognition Method Based on Energy Spectrum Analysis and DBSCAN Improved Algorithm

作　　者：万中一 胡岳[1,2] 陈炳树[1,2] 李延栋 WAN Zhongyi;HU Yue;CHEN Bingshu;LI Yandong(Department of Electrical Engineering,School of Electronic Information and Electrical Engineering,Shanghai Jiao Tong University,Shanghai 200240,China;Key Laboratory of Control of Power Transmission and Conversion(Shanghai Jiao Tong University),Ministry of Education,Shanghai 200240,China;State Grid Gansu Electric Power Company,Lanzhou 730070,China)

机构地区：[1]上海交通大学电子信息与电气工程学院电气工程系,上海200240 [2]电力传输与功率变换控制教育部重点实验室(上海交通大学),上海200240 [3]国网甘肃省电力公司,兰州730030

出　　处：《高电压技术》2023年第10期4335-4344,共10页High Voltage Engineering

基　　金：国家自然科学基金(51777122)。

摘　　要：针对带电检测现场存在多源局部放电源及干扰脉冲信号,不能直接获得单类局部放电相位分布(phase resolved partial discharge,PRPD)谱图进行模式识别的问题,该文提出基于能量谱分析和基于密度的噪声应用空间聚类(density-based spatial clustering of applications with noise,DBSCAN)改进算法的局部放电多源信号分类识别方法。首先,采用能量谱分析提取检测脉冲选定频段的能量百分占比作为局部放电时间分布(time resolved partial discharge,TRPD)谱图特征值,该选定频段的中心频率和带宽由传感器性能参数确定;进而,根据单类脉冲信号选定频段的能量百分占比密度近似正态分布,提出并采用改进的DBSCAN算法实现多源信号的自适应聚类分离;最后,基于分类信号的PRPD谱图实现了局部放电信号的类型识别。实验结果表明,多源局部放电信号的识别准确率达到98.45%,该方法可用于变电站多源局部放电的现场带电检测与分类识别。Aiming at the problem that there are multi-source partial discharge power sources and interference pulse signals in live detection field,and the single-type phase resolved partial discharge(PRPD)spectrum cannot be directly obtained for pattern recognition,we put forward a classification and identification method of multi-source signals based on energy spectrum analysis and density-based spatial clustering of applications with noise(DBSCAN)improved algorithm.Firstly,the energy ratio of the selected frequency band is extracted by energy spectrum analysis and is adopted as the time resolved partial discharge(TRPD)characteristic values,and the center frequency and bandwidth of the selected frequency band are determined by the sensor performance parameters.Furthermore,according to the approximate normal distribution of the energy percentage density of the selected frequency band in the single pulse signal,an improved DBSCAN algorithm is proposed to realize the adaptive clustering and separation of multi-source signals.Finally,partial discharge type identification is realized based on the PRPD spectrum of classified signals.The experimental results show that the recognition accuracy of multi-source partial discharge signal reaches 98.45%,and this method can be used for live detection and classification of multi-source partial discharge in substation.

关 键 词：局部放电 多源信号 能量谱 模式识别 DBSCAN改进算法 

分 类 号：TM855[电气工程—高电压与绝缘技术]