支持向量机在电力变压器故障诊断中的应用

利用支持向量机的学习方法，构建了电力变压器故障诊断模型．该模型将变压器故障分为放电性和过热性两大类，通过统计分析寻求特征量区分类间的故障类型，采用支持向量机识别类内的故障类型，利用基于交叉验证的网格搜索法来确定支持向量机的参数．考虑到变压器油中溶解气体特征空间的紧致性原理，利用模糊C均值聚类算法对所获取的样本进行预选取，有效地解决了确定模型参数时耗时巨大的问题，并一定程度提高了模型的推广能力．实例验证表明，该模型在有限样本情况下，能达到较高的变压器故障判断率，放电性故障样本正确判断率为90．5％，过热性故障样本正确判断率为85．9％，说明该模型具有很好的分类效果和推广能力．

Application of Support Vector Machine in Transformer Fault Diagnosis

A model of transformer diagnosis based on support vector machine is proposed, where transformer faults are divided into two types, discharge fault and thermal fault, and the two fault types are recognized with their statistical features. The support vector classifier is adopted to identify fault in the two types, and the grid search method based on cross-validation is chosen to determine model parameters. Considering the compactness characteristics of dissolved gas analysis data, the achieved samples are pre-selected with the fuzzy C-means clustering method to solve the problem of long time consuming in parameter determination, thus a certain model extension ability is enhanced. The experiment shows that the model enables to detect transformer faults with a higher diagnosis rate, under condition of small samples, the diagnosis rate for discharge fault samples gets 90.5%, and 85.9% for thermal fault samples.