摘要
基于油纸绝缘柱板沿面放电模型,采用恒压法对该模型交流电压下局部放电发展特性进行实验研究,提取不同时刻局部放电脉冲相位分布(phase-resolved partial discharge,PRPD)模式的4个图谱及其对应的29个统计特征量,采用核主成分分析法提取8个新特征参量,并通过系统聚类划分放电发展阶段。在局部放电8个新特征参量的基础上,根据聚类划分的结果,利用聚类分析中的类距离思想,建立基于聚类-多级支持向量机(support vector machine,SVM)的放电阶段分层识别方法。识别结果表明:所建立的聚类-多级SVM能很好识别放电所处的阶段,与系统聚类结果吻合较好。
Based on surface discharge model for oil paper insulation rod-plane,constant voltage method was adopted for experiment research on development characteristic of partial discharge( PD) of this model under AC voltage. Four graphs of phase-resolved partial discharge( PRPD) modes at different time and 29 statistic characteristic quantities were extracted,kernel principle component analysis method was used for extracting eight newcharacteristic quantities and discharge development stages were divided by system clustering analysis as well. On the basis of eight newcharacteristic parameters,class distance point in clustering analysis was used to form a hierarchical identification method for discharge stages based on cluster-multistage support vector machine( SVM). Results indicate this method based on cluster-multistage SVMcan identify discharge stages well and get result consistent with that of system clustering analysis method.
出处
《广东电力》
2017年第9期104-110,共7页
Guangdong Electric Power
基金
国家自然科学基金创新群体基金项目(51321063)
中国南方电网有限责任公司重大科技专项(KJ-2014-170-3)
关键词
油纸绝缘
局部放电
发展特性
支持向量机
阶段识别
oil paper insulation
partial discharge
development characteristic
support vector machine
stage identification
