相关向量机及其在变压器故障诊断中的应用

摘要

分析并用典型数据分类算例验证相关向量机(RVM)在分类性能方面优于支持向量机(SVM),在此基础上以标准化的变压器主要特征气体含量为输入量,采用二叉树的分类方法建立基于RVM的变压器故障诊断模型.实例分析表明,同基于SVM的故障诊断方法相比,该方法可以取得与其相当甚至更优的故障诊断正确率,相关向量个数明显少于支持向量个数,诊断速度显著提高.%Analysis and typical data classification examples validate the classification performance of RVM (Relevance Vector Machine) is better than that of SVM (Support Vector Machine). A transformer fault diagnosis method based on RVM is put forward, which takes the normalized contents of transformer feature gases as inputs and adopts the binary tree classification means. Experimental results show that, compared with the fault diagnosis method based on SVM,it gets comparable or better diagnostic accuracy with less vector amount and faster diagnosis speed.