基于支持向量机的时序周波波形分类方法  

Support vector machine based classification method for time-series periodic waveform

作　　者：胡志坤[1] 王美铃[1] 桂卫华[2] 阳春华[2] 丁家峰[1] 

机构地区：[1]中南大学物理科学与技术学院,湖南长沙410083 [2]中南大学信息科学与工程学院,湖南长沙410083

出　　处：《浙江大学学报（工学版）》2010年第7期1327-1332,共6页Journal of Zhejiang University：Engineering Science

基　　金：国家自然科学基金资助项目(60904077;60874069);深圳市科技计划基础研究资助项目(JC200903180555A)

摘　　要：针对电力系统输出的周波波形多的特点,提出一种基于小波分析和支持向量机(SVM)的时序周波波形分类方法,实现三相电压源型逆变器的故障分类.利用离散正交小波变换(DOWT)将周波序列变换成小波系数矩阵,利用奇异值分解(SVD)的方法获得系数矩阵的奇异值向量,作为周波序列的特征值.建立基于新的Huffman树来实现支持向量机策略的多类分类模型.将奇异值分解得到的特征向量应用到该分类模型,判断逆变器的故障类型.仿真结果表明,该模型的平均期望准确率比基于普通二叉树的支持向量机多类模型高3.65%,分类准确率达到99.6%.Aimed at the characteristic of power system possessing lots of periodic waveforms,a new classification method for time-series periodic waveform was proposed based on the wavelet analysis and the support vector machine（SVM） in order to realize the fault type classification of three-phase voltage inverter.The periodic waveform was transformed into the wavelet coefficient matrix by using the discrete orthogonal wavelet transformation（DOWT）.Then the singular value vector was obtained using the singular value decomposition（SVD） method,and acted as the feature value of time-series periodic waveform.A multi-classes classification model was established based on a new Huffman tree in order to realize the SVM strategy.The extracted feature vectors were applied to the classification model in order to judge the fault type of the inverter.Simulation results showed that the average Loo-correctness of the model exceeded the ordinary binary tree based SVM 3.65%,and the correctness of classification reached 99.6%.

关 键 词：小波变换 时序序列 支持向量机(SVM) 故障诊断 

分 类 号：TM401[电气工程—电器]