基于振荡波多特征融合的变压器绕组故障诊断方法

Transformer winding fault diagnosis method based on oscillating wave multi-feature fusion

为了获取更多变压器绕组的状态信息,提高绕组故障诊断的准确性,提出了一种基于振荡波多特征融合的变压器绕组故障诊断方法,该方法联合波形特征和小波包时频图的颜色特征判断故障类型、故障程度和故障位置,结合粒子群优化-支持向量机(PSO-SVM)算法实现变压器绕组状态的智能识别。最后搭建变压器故障模拟试验平台验证方法可行性。结果表明:波形特征、颜色矩、颜色聚合向量特征分别针对故障类型、故障程度及故障位置的空间分布具有分离和聚类特性,且通过PSO-SVM识别的准确率高达95%以上,故所提方法能够准确辨识变压器绕组的状态,为现场变压器绕组状态检测提供参考。

In order to obtain more state information of transformer windings and improve the accuracy of winding fault diagnosis,a transformer winding fault diagnosis method based on oscillating wave multi-feature fusion is proposed.In this method,the fault type,fault degree and fault location are judged by the waveform characteristics of oscillating wave and the colour characteristics of time-frequency graph,and the automatic recognition of transformer winding state is realized by PSO-SVM(Particle Swarm Optimization-Support Vector Machine).The proposed method is verified by transformer fault simulation test platform.The results show that waveform feature,colour moment and colour coherence vector have the characteristics of separation and clustering respectively for the spatial distribution of fault type,fault degree and fault location,and the recognition accuracy of PSO-SVM is higher than 95%,so the proposed method can accurately identify the state of transformer winding,and provide a reference for on-site transformer winding status detection.