基于分步特征选取和WOA-LSSVM的变压器故障诊断

Transformer Fault Diagnosis Based on Stepwise Feature Selection and WOA-LSSVM

为了提高变压器故障诊断的精度,保障电网的稳定运行,提出了一种基于ReliefF算法与界标等距映射(L-Isomap)的分步特征选取和鲸鱼群算法(WOA)优化最小二乘支持向量机(LSSVM)的故障诊断模型。选取7种常见故障特征油中溶解气体分析(DGA)气体以及其构造出的16组比值作为初始特征集,利用ReliefF算法分别对初始特征集进行特征选择,再利用L-Isomap算法对融合后的特征集进行降维处理,将降维处理后的特征集作为故障特征向量代入诊断模型,故障诊断模型采用WOA-LSSVM进行训练与测试。实验结果表明,诊断模型的精度高达98.31%,相比于其他模型拥有更高的诊断精度。

In order to improve the accuracy of transformer fault diagnosis and ensure the stable operation of power system.In this paper,proposing a stepwise feature selection based on ReliefF algorithm and landmark isomap(L-Isomap)and a fault diagnosis model for whale optimization algorithm(WOA)least squares support vector machine(LSSVM).The method first selected 7 common fault characteristics dissolved gas analysis in oil(DGA)gas and constructed 16 sets of ratios as the initial feature set.Secondly,the ReliefF algorithm was used to perform feature selection on the initial feature set respectively,and then the L-Isomap algorithm was used to reduce the dimensionality of the fused feature set,and the dimensionality reduction feature set was substituted into the diagnostic model as a fault feature vector,and the fault diagnosis model was trained and tested by WOA-LSSVM.The experimental results show that the accuracy of the diagnostic model is as high as 98.31%,which is higher diagnostic accuracy than that of other models.