摘要
支持向量机最初只能解决二分类问题,在解决故障诊断多分类问题时存在较大困难。在研究二叉树多分类的基础上,结合变压器故障的特点,运用相关性方法得到故障之间的近似程度,针对变压器多故障建立了一种改进的决策树模型。由于训练样本的不平衡性,对支持向量机的惩罚系数进行了优化,这样大大避免了训练样本带来的模型缺陷问题。根据变压器现场实际采集故障数据测试结果表明,改进的决策树分类器模型针对变压器故障诊断具有优越的应用价值。
Support Vector Machine is a two classifiers,There are greater difficulties to solve the fault diagnosis of multi-classification problem.Combined with the characteristic of the transformer failure and using the method of correlation failure between the degree of approximation,the multiclass SVM methods based on decision tree are proposed.Taking into account the imbalance of the training sample,the penalty of support vector machine has been improved soasto effectively avoid the model from the training sampledefects.
出处
《电子质量》
2015年第8期88-92,共5页
Electronics Quality
关键词
支持向量机
故障诊断
决策树
变压器
Support Vector Machine
fault diagnosis
decision tree
classification
