基于油中气体分析与类重叠特征的变压器分层故障诊断模型  

Transformer Hierarchical Fault Diagnosis Model Based on Dissolved Gas Analysis of Insulating Oil and Class Overlap Features

作　　者：陈铁[1,2] 冷昊伟 李咸善[1,2] 陈一夫 CHEN Tie;LENG Haowei;LI Xianshan;CHEN Yifu(Hubei Provincial Key Laboratory for Operation and Control of Cascaded Hydropower Station,China Three Gorges University,Yichang 443000,China;School of Electrical Engineering and New Energy,China Three Gorges University,Yichang 443000,China.)

机构地区：[1]三峡大学水电站运行与控制湖北省重点实验室,湖北宜昌443000 [2]三峡大学电气与新能源学院,湖北宜昌443000

出　　处：《中国电力》2022年第7期22-32,41,共12页Electric Power

基　　金：国家自然科学基金资助项目(51741907);梯级水电站运行与控制湖北省重点实验室开放基金(2019KJX08)。

摘　　要：油中溶解气体分析可以有效识别变压器放电故障与过热故障,为提高变压器故障诊断准确度,提出一种基于类重叠特征的变压器分层故障诊断方法。首先使用支持向量数据描述(SVDD)划分出变压器故障样本数据空间的重叠区域,选择类重叠率与类重叠度作为重叠特征,分别对类重叠程度和样本点重要性进行描述,然后以类重叠率为分层标准建立分层故障诊断模型,采用分隔训练法将各诊断层的样本集分开训练,针对分类难度较大的重叠区,基于类重叠度构造二分类模糊支持向量机(FSVM)进行故障诊断。实验结果表明,相比于其他模型,所提方法具有更高的准确度。Dissolved gas analysis(DGA)of insulating oil can effectively identify transformer discharge fault and overheating fault.In order to improve the accuracy of transformer fault diagnosis,a transformer hierarchical fault diagnosis method is proposed based on class overlap features.Firstly,the support vector data description(SVDD)is used to divide the overlapping region of transformer fault sample data spaces,and the class overlap rate and class overlap degree are selected as the overlapping features to describe the class overlap degree and the importance of sample points respectively.And then,a hierarchical fault diagnosis model is established based on the class overlap rate.The samples of each diagnosis layer are trained separately by the separate training method,and a twoclass fuzzy support vector machine(FSVM)is constructed based on class overlap degrees to diagnose faults.Experimental results show that the proposed method is more accurate than other models.

关 键 词：变压器故障诊断 类重叠 分层诊断 支持向量数据描述 模糊支持向量机 

分 类 号：TM41[电气工程—电器]