基于特征优选和HHO-SVM的变压器故障识别

Transformer Fault Recognition Based on Feature Selection and HHO-SVM

在油中溶解气体的分析方法的研究中,不同故障类型下,各气体所体现的故障特征存在细微的区别。依次对不同故障的数据提取特征,能够补充和完善变压器故障特征的表征。数据集包括5种气体含量比值和三比值数集,将其拆分为6种不同故障类型的子集,利用核主元分析每一种故障类型的主元特征量,综合所有故障特征量,通过哈里斯鹰搜索算法,优化SVM模型进行故障识别验证,准确率为94.1%。实例结果表明,该方法保留了不同的故障类型的特征差异,提升了特征样本集的可分性。

In the study of the analysis method of dissolved gas in oil,the fault characteristics of each gas are slightly different under different fault types.The feature extraction of different fault types can supplement and improve the characterization of transformer fault characteristics.According to the actual fault of transformer operation.The data set includes five kinds of gas content ratio and three ratio number set,which are divided into six different fault type subsets.The kernel principal component analysis model is used to analyze the principal component features of each fault type,and all fault features are integrated.The Harris hawk search algorithm is used to optimize the SVM model for fault recognition verification.The accuracy is 94.1%.The results of case study show that the method retains the feature differences of different fault types and makes the separability of feature sample set larger.