摘要
针对现有支持向量机诊断模型构造复杂、参数设计困难等问题,建立了基于多类支持向量机的变压器故障诊断模型,并提出一种实用的支持向量机参数寻优方法。该方法结合网格搜索,对训练样本进行分组交叉验证寻找给定范围内的最优参数,有效地解决了支持向量机的参数设计难题。实例计算表明,基于参数寻优建立的多类支持向量机模型在保证很高的故障分类正确率的同时,大大降低了二值支持向量机分层组合模型的构造及参数设计的复杂程度,具有很好的实用性和推广性。
In allusion to the disadvantages of traditional SVM （Support Vector Machine） classification model, a transformer fault diagnosis model based on multi-class SVM is constructed. As an important step, a practical parameter optimization method for SVM classifier which is based on ＂Grid-search＂and ＂Cross-validation＂algorithm is presented then. The optimum parameters are achieved through an approach which validates the SVM models by grouping training data sets;meanwhile, the parameters of each SVM model are changed following grid-search ways. Compared with the classification model based on layered combined binary SVMs, the multi-class SVM classification model with optimum parameters in this paper has high classification accuracy in the same and, moreover, several advantages such as conveniences of model construction and parameters selection. The effectiveness and usefulness of this model is proved by a practical example.
出处
《水电能源科学》
2007年第1期52-55,共4页
Water Resources and Power
作者简介
江伟（1983-），男，硕士研究生，研究方向为电气设备状态维护与故障诊断，E-mail：inorthstar@163．com
通讯作者：罗毅（1966-），男，副教授，研究方向为EMS和DMS，E-mail：luoyee@mail．hust．edu．cn
