摘要
针对将神经网络应用于电力变压器故障诊断时输入矢量的选择问题，提出了一种基于多元统计分析技术的输入矢量选择方法，利用聚类分析和因子分析技术对油中溶解气体的结果进行了处理，并对输入矢量处理前、后的反向传播神经网络的分类能力进行了比较，结果表明该方法有效。
In order to acquire the suitable input vectors ofneural network applied in fault diagnosis of power transformer, a method based on the Multivariate StatisticalAnalysis(MSA) is presented in this paper. The results ofProcessed Dissolved Gas Analysis (DGA) in oil are treatedby the technique of cluster analysis and factor analysis. Andthe results show that the classifying ability of Back Propagation Neural Network (BPNN ) is improved by thismethod.
出处
《高电压技术》
EI
CAS
CSCD
北大核心
1999年第2期6-8,12,共4页
High Voltage Engineering
基金
国家自然科学基金
东北电力集团资助项目!59637200
关键词
电力变压器
过热
放电
故障诊断
多元统计分析
power transformer back propagation neuralnetwork dissolved gas analysis fault diagnosis multivariate statistical analysis
