摘要
针对变压器故障诊断中缺少实际典型故障样本的问题,提出了支持向量机(SVMs)变压器故障诊断方法。该方法采用K均值聚类(KMC)对变压器油中5种特征气体样本进行预选取作为特征向量,输入到多分类支持向量机中进行训练,建立SVMs诊断模型,实现对故障样本的诊断分类。实例分析表明,KMC算法浓缩了故障信息,有效地解决了确定模型参数时耗时巨大的问题。该方法在有限样本情况下,能够达到较高的故障正判率,满足变压器故障自动诊断的目的。
Due to lack of typical damage samples in the transformer fault diagnosis,a new fault diagnosis method based on support vector machines(SVMs) is presented.According to the method,the five characteristic gases dissolved in transformer oil are extracted by the K-means clustering(KMC) method as feature vectors,which are input into multi-classified SVMs for training,and then the SVMs diagnosis model is established to implement fault samples classification.The results of experiment and analysis show that with KMC algorithm,the diagnosis information are concentrated and the great time consumption in parameter determination is remitted effectively.The presented method can detect the faults in transformer with a high correct judgment rate and can reach the purpose of automation diagnosis for transformer faults under the condition of few samples.
出处
《现代电子技术》
2011年第24期118-120,共3页
Modern Electronics Technique
基金
国家科技支撑计划项目:高大空间建筑工程安装维护设备技术与产业化开发(2008BAJ09B06)
中国博士后基金:斜拉桥损伤识别和健康状态预测技术研究(20110491637)
关键词
变压器
故障诊断
K均值聚类
支持向量机
transformer
fault diagnosis
K-means clustering
support vector machine