基于ReliefF和HPO-SVM的变压器故障检测方法  

Transformer Fault Diagnosis Method Based on ReliefF and HPO-SVM

作　　者：张晓虎 宁环宇 ZHANG Xiaohu;NING Huanyu(College of Electrical and Information Engineering,Hunan University of Technology,Zhuzhou 412000,China)

机构地区：[1]湖南工业大学电气与信息工程学院,湖南株洲412000

出　　处：《电工技术》2023年第17期1-5,共5页Electric Engineering

基　　金：国家重点研发计划项目(编号2022YFE0105200)。

摘　　要：为提高油浸式电力变压器故障诊断的判断正确率,提出了一种利用ReliefF特征权重法、HPO-SVM模型和油中溶解气体分析法(DGA)相结合的故障诊断方法。首先,该方法引入特征权重算法对输入量进行筛选降维;其次,采用猎食者优化算法对概率神经网络模型进行了优化,利用SVM模型处理DGA比值集合,最终得到变压器的故障诊断结果。实验结果表明,采用ReliefF特征权重算法进行降维的模型拥有更高的诊断精确度。实验结果证明HPO-SVM、GWO-SVM、WOA-SVM、PSO-SVM的平均故障判断准确率分别为94%、91.33%、90%、83.33%。仿真结果表明,优选后的混合特征模型诊断正确率更高,证实了此方案的优越性。In order to improve the accuracy of fault diagnosis of oil-immersed power transformer,a fault diagnosis method based on the integration of ReliefF feature weight method,HPO-SVM model and dissolved gas analysis in oil(DGA)is proposed in this paper.Firstly,the feature weight algorithm is employed to filter the input dimension.Secondly,the probabilistic neural network model is optimized by using the HPO algorithm,and the DGA ratio set is processed by the SVM model.Finally,the transformer fault diagnosis results are obtained.Experiments show that dimension reduction by the ReliefF feature weight algorithm has higher diagnosis accuracy.The average fault diagnosis accuracy of HPO-SVM,GWO-SVM,WOA-SVM,and PSO-SVM is 94%,91.33%,90%,and 83.33%,respectively.In addition,the simulation shows that the hybrid feature model sifted out has higher rate of correctness in fault diagnosis.The results confirmed the superiority of the proposed scheme.

关 键 词：变压器故障诊断 油中溶解气体分析 特征权重 猎食者优化支持向量机 

分 类 号：TM411[电气工程—电器]