基于特征评估与核主元分析的电力变压器故障诊断
发布时间：2019-02-25 21:15
【摘要】：针对电力变压器故障诊断中的故障特征量数量匮乏、携带的故障信息较为有限,致使故障判断效果不理想等问题,将电气试验数据等与油中溶解气体分析(DGA)相融合所获得的34种特征量作为故障特征量,以完善故障特征信息。在此基础上,将特征评估与核主元分析(KPCA)相结合,构建了一种基于特征评估与核主元分析的故障诊断方法。该方法首先通过特征评估来剔除不敏感故障特征量,以削弱它们对特征提取产生的影响;其次,对经过特征评估后的27维故障特征量进行核主元分析,降低故障特征量的维数;最后,将提取后的9维故障特征量作为输入故障特征向量,采用多分类相关向量机(M-RVM)方法进行故障分类。实例分析表明,该故障诊断方法不仅能有效弥补故障特征量单一等不足,而且更具一般性,其故障诊断准确率达到90.35%,可为故障信息有限情况下的电力变压器故障诊断提供参考。
[Abstract]:In view of the shortage of fault characteristic quantity in fault diagnosis of power transformer and the limited fault information carried by it, the result of fault diagnosis is not satisfactory and so on. 34 kinds of characteristic parameters obtained from the fusion of electrical test data and dissolved gas analysis (DGA) in oil are used as fault characteristics to improve the fault characteristic information. On this basis, a fault diagnosis method based on feature evaluation and kernel principal component analysis (KPCA) is proposed by combining feature evaluation with kernel principal component analysis (KPCA). Firstly, the insensitive fault features are eliminated by feature evaluation to weaken their influence on feature extraction, secondly, the kernel principal component analysis is carried out to reduce the dimension of fault feature variables after the 27-dimensional fault feature analysis after feature evaluation. Finally, the 9-dimensional fault feature is used as the input fault feature vector, and the multi-classification correlation vector machine (M-RVM) is used to classify the fault. The example analysis shows that the fault diagnosis method can not only make up for the deficiency of single fault characteristic quantity, but also has more generality. The accuracy of fault diagnosis is 90.35%, and the fault diagnosis accuracy is 90.35%. It can provide reference for power transformer fault diagnosis when fault information is limited.
【作者单位】： 西南交通大学电气工程学院;
【基金】：国家自然科学基金(U1234202) 国家杰出青年基金(51325704)~~
【分类号】：TM41
本文编号：2430544
[Abstract]:In view of the shortage of fault characteristic quantity in fault diagnosis of power transformer and the limited fault information carried by it, the result of fault diagnosis is not satisfactory and so on. 34 kinds of characteristic parameters obtained from the fusion of electrical test data and dissolved gas analysis (DGA) in oil are used as fault characteristics to improve the fault characteristic information. On this basis, a fault diagnosis method based on feature evaluation and kernel principal component analysis (KPCA) is proposed by combining feature evaluation with kernel principal component analysis (KPCA). Firstly, the insensitive fault features are eliminated by feature evaluation to weaken their influence on feature extraction, secondly, the kernel principal component analysis is carried out to reduce the dimension of fault feature variables after the 27-dimensional fault feature analysis after feature evaluation. Finally, the 9-dimensional fault feature is used as the input fault feature vector, and the multi-classification correlation vector machine (M-RVM) is used to classify the fault. The example analysis shows that the fault diagnosis method can not only make up for the deficiency of single fault characteristic quantity, but also has more generality. The accuracy of fault diagnosis is 90.35%, and the fault diagnosis accuracy is 90.35%. It can provide reference for power transformer fault diagnosis when fault information is limited.
【作者单位】： 西南交通大学电气工程学院;
【基金】：国家自然科学基金(U1234202) 国家杰出青年基金(51325704)~~
【分类号】：TM41
【相似文献】
相关期刊论文 前2条
1 郑育平;张丽萍;;基于核主元分析的湿法烟气脱硫系统的故障诊断[J];福州大学学报(自然科学版);2013年03期
2 李平;李学军;蒋玲莉;曹宇翔;;基于KPCA和PSOSVM的异步电机故障诊断[J];振动.测试与诊断;2014年04期
，本文编号：2430544
本文链接：https://www.wllwen.com/kejilunwen/dianlidianqilunwen/2430544.html
