高电压技术2023，Vol.49Issue(12)：4993-5001,9.DOI:10.13336/j.1003-6520.hve.20221528
不平衡样本下基于变异麻雀搜索算法和改进SMOTE的变压器故障诊断方法
Transformer Fault Diagnosis Method Based on Variation Sparrow Search Algorithm and Improved SMOTE Under Unbalanced Samples
摘要
Abstract
To solve the problem of poor classification results caused by the serious homogeneity of the sparrow search algorithm and the unbalance of the transformer fault sample,a method of transformer fault diagnosis is proposed.It is based on a variation sparrow search algorithm-support vector machine(VSSA-SVM)and improved synthetic minority over-sampling technique(ISMOTE).Firstly,the dataset is denoised by the Tomek Link,and the center offset weight(COW)is introduced to improve the SMOTE for minority class sample synthesis of the unbalanced dataset,which can obtain the balancing processed transformer fault dataset.Then,a transformer fault diagnosis model based on the variation of the VSSA-SVM is proposed using the idea of mutation.Finally,the dissoived gas anaiysis(DGA)data of 413 cases of the collecting oil-immersed transformers are diagnosed using SSA-SVM,PSO-SVM and VSSA-SVM models,and the diagnosis accuracy results are 81.45％,88.71％and 96.77％respectively.Additionally,compared to the SMOTE-NND,SVM SMOTE,Borderline-SMOTE,SMOTE and original dataset methods,the diagnosis accuracy of the ISMOTE model is improved by 3.22％,4.03％,6.45％,7.52％and 11.29％respectively.The results show that the proposed method in this paper can be adopted to accurately judge the transformer fault state,and to effectively solve the problem of unbalanced fault data caused by low classification accuracy,which has a certain engineering practical value.关键词
变压器/故障诊断/不平衡样本/改进合成少数过采样/变异麻雀搜索算法Key words
transformer/fault diagnosis/unbalanced samples/improved synthetic minority oversampling/variation sparrow search algorithm引用本文复制引用
朱莉,汪小豪,李豪,姜成龙,曹明海..不平衡样本下基于变异麻雀搜索算法和改进SMOTE的变压器故障诊断方法[J].高电压技术,2023,49(12):4993-5001,9.基金项目
新能源及电网装备安全监测湖北省工程研究中心开放研究基金(HBSKF202124).Project supported by New Energy and Grid Equipment Safety Monitoring Hubei Province Engineering Research Center Open Research Fund(HBSKF202124). （HBSKF202124）
