摘要
针对变压器多故障问题,提出了基于Mercer核函数的欧式距离查询策略算法,并建立了基于Karhunen-Loeve(K-L)特征提取与支持向量机的变压器故障诊断模型,利用K-L变换提取信号的特征值,最后通过支持向量机学习算法完成对信号的选择与分类。通过实例应用表明:所训练的SVM分类器较之直接任意选取训练样本作为训练集的传统方法具有更高的诊断率。
Aiming at the multi - fault problems of transformer, the Euclidian distance query algorithm based on Mercer kernel function is proposed and a model based on Karhunen - Loeve （ K - L） feature extracting and support vector machine （SVM） is established for fault diagnosis of oil - immersed transformer, the eigenvalue of signal is extracted using K - L transform, and finally, the SVM learning algorithm is introduced to select and classify the training sample data. The result shows that the precision of the trained SVM classifier is better than that of the traditional method, and the reliability and effectiveness using the above mentioned method are satisfied in fault diagnosis.
出处
《四川电力技术》
2012年第6期58-61,共4页
Sichuan Electric Power Technology
关键词
油浸变压器
故障诊断
支持向量机
K—L特征提取
oil - immersed transformer
fault diagnosis
support vector machine
K - L feature extracting
