基于组合模型的电力变压器故障诊断

Fault Diagnosis of Power Transformer Based on Combined Model

为了综合全面地诊断电力变压器故障,克服单项诊断方法考虑问题角度单一,不能重复利用已知信息,诊断准确度和稳定性不高的缺点,并结合电力变压器油中溶解气体的数据,提出了利用组合模型诊断变压器故障.该方法将灰关联熵、小波神经网络、模糊粗糙集、支持向量机和IEC三比值作为独立诊断模块,利用熵值法优化得到各个模块的最佳权重,最终得到发生故障最大概率所属类型.通过实例验证,组合诊断法优于单项诊断方法,提高了故障诊断精度,减少了误判率,诊断的稳定性得到提升.

In order to comprehensively diagnose transformer faults and overcome one single fault diagnostic methodwith low diagnostic accuracy and the low stability,a new way,which composites several transformer fault diagnosismethods,is brought forward to overcome the limitations of a single method. Grey relational entropy,wavelet neuralnetwork,fuzzy rough sets,support vector machine,three-ratio method are adopted as independent methods to diagnosefaults in the new diagnosis model. According to diagnostic accuracy,the optimal value weights are got with the entropyoptimization. Then the optimal combination model is set up and the fault type with the largest fault possibility isacquired. The results of diagnosis instances show that the new method not only decreases misjudgment risk and getsmuch higher accuracy rate,but also enhances diagnostic robustness compared with single method.