基于谱聚类集成的变压器在线故障诊断

发布时间：2018-04-09 15:10

本文选题：故障诊断　切入点：变压器　出处：《电子学报》2017年10期

【摘要】：为了提高基于油中溶解气体分析技术(DGA)的变压器故障诊断准确率,本文提出了一种基于谱聚类集成的变压器在线故障诊断(TOFD-SCE)方法.以加权二次抽样算法抽取样本、构建基础谱聚类的样本集,以基础谱聚类学习问题的局部知识;平衡多样性与正确性选择集成成员;集成多个成员谱聚类的结果来提高变压器故障诊断的准确率.传统变压器故障诊断方法基于历史数据建立模型,缺乏在线学习的能力;TOFD-SCE利用历史与在线新增两种DGA数据来训练、修正模型,提高了其故障诊断的准确率.对SSP300000/500型变压器的故障诊断实验结果表明:TOFD-SCE的准确率优于IEC三比值法、BP神经网络法及支持向量机法,验证了其有效性.
[Abstract]:In order to improve the accuracy of transformer fault diagnosis based on dissolved gas analysis in oil (DGA), a method of on-line fault diagnosis of transformer based on spectral clustering integration is proposed in this paper.The weighted quadratic sampling algorithm is used to extract samples and construct the sample set of the basic spectral clustering, which is used to learn the local knowledge of the basic spectral clustering problem, and to balance diversity and correctness to select the integrated members.The accuracy of transformer fault diagnosis is improved by integrating the results of multiple member spectrum clustering.The traditional method of transformer fault diagnosis is based on historical data and lacks the ability of on-line learning. TOFD-SCE uses historical and online new two kinds of DGA data to train and modify the model to improve the accuracy of fault diagnosis.The experimental results of fault diagnosis for SSP300000/500 type transformer show that the accuracy of 1: TOFD-SCE is better than that of IEC three-ratio method, BP neural network method and support vector machine method, and the validity of the method is verified.
【作者单位】：湖南大学电气与信息工程学院;长沙理工大学电气与信息工程学院;
【基金】：国家自然科学基金(No.61472128,No.61173108) 湖南省自然科学基金重点项目(No.14JJ2150) 国家电网公司总部科技项目(No.5216A514001K)
【分类号】：TM41

【相似文献】