基于支持向量机的分类辨识方法及应用  

A Classification Method Based on Support Vector Machine and Its Applications

作　　者：马相东 卢占庆 谭永彦 王秀英[3] 

机构地区：[1]山东省冶金设计院股份有限公司,山东莱芜271104 [2]山东钢铁股份有限公司莱芜分公司,山东莱芜271104 [3]青岛科技大学信息科学技术学院,山东青岛266061

出　　处：《控制工程》2016年第5期768-772,共5页Control Engineering of China

基　　金：国家自然科学基金(61104004);山东省自然科学基金(ZR2014FL019);青岛市科技计划(13-1-4-125-jch)

摘　　要：针对钢铁企业传统高压线路行波检测器易出现误报警的问题,提出了采用支持向量机(SVM)的分类辨识方法,用于故障信号和扰动信号的辨识。首先,从行波法定位装置数据库中提取报警时的信号特征,构造支持向量的输入向量,建立基于数据驱动的支持向量机模型;然后采用仿真策略确定支持向量机径向基函数中的参数σ和惩罚系C的值,并分析了参数σ及惩罚系数C的值对故障和扰动分类准确率的影响。将所提出的方法应用到莱钢高压线路行波检测器中,结果表明:采用支持向量机的分类辨识方法,可以使行波检测器检测的准确率接近90%,大大提高了莱钢高压线路行波检测器故障检测的可靠性。A classification method based on support vector machine（SVM） is proposed to distinguish disturbance from failure on traveling wave failure detection device for traditional high-voltage circuit in metal industry.Initially recorded data for fault and disturbance are selected from database,and then input features and output feature are extracted,which are used for data driven classification model.After that,the model is established,and its parameters,i.e.,parameters of the radial basis function and the penalty factor C,are optimized by MATLAB simulation method.In addition,an analysis on influence of selected model parameters to model classification accuracy is also carried out.The proposed method is applied to traveling wave detection device in Laiwu Iron Steel Company,and the classification accuracy reaches 90%,which enhances the reliability of failure detection compared to the traditional method and achieves the expected results.

关 键 词：行波法故障检测 支持向量机 故障信号 扰动信号 分类辨识 

分 类 号：TP29[自动化与计算机技术—检测技术与自动化装置]