综合三比值特征量与帝国竞争优化支持向量机的变压器故障诊断模型

The Fault Diagnosis Model of Power Transformer Fusing the Three DGA Ratios and the ICA Optimized SVM

为了提高变压器故障诊断准确率,提出了一种综合三比值特征量与帝国竞争优化(ICA)支持向量机(SVM)的变压器故障诊断模型。该模型将三比值特征量作为输入,采用基于“留一法”的平均分类准确率构建目标函数,通过ICA优化SVM对变压器进行故障诊断。实例分析结果显示:该模型相比标准SVM法和BP神经网络法,准确率提高了6%~34%,其训练样本平均准确率和测试样本平均准确率分别达到了91.5%和86.7%。

In order to improve the accuracy of transformer fault diagnosis,a transformer fault diagnosis model fusing the characteristic quantity of three DGA ratios and the imperialist competition algorithm(ICA) optimized support vector machine(SVM) is proposed.Firstly,the three DGA ratios are used as the input vector in the model;then the objective function is constructed by the average classification accuracy by using the‘leave-one-out method’(LOOM).Finally,the transformer fault is diagnosed by the ICA optimized SVM.The analysis results of practical examples show that the accuracy rate of the model is 6% to 34% higher than that of the standard SVM method and BPNN method and the average accuracy rate of training samples and testing samples are 91.5% and 86.7% respectively.