Please wait a minute...
浙江大学学报(工学版)
J4  2010, Vol. 44 Issue (7): 1327-1332    DOI: 10.3785/j.issn.1008-973X.2010.07.017
自动化技术     
基于支持向量机的时序周波波形分类方法
胡志坤1, 王美铃1, 桂卫华2, 阳春华2, 丁家峰1
1.中南大学 物理科学与技术学院,湖南 长沙 410083；中南大学 信息科学与工程学院,湖南 长沙 410083
Support vector machine based classification method for
timeseries periodic waveform
HU Zhikun1, WANG Meiling1, GUI Weihua2, YANG Chunhua2, DING Jiafeng1
1.School of Physics Science and Technology, Central South University, Changsha 410083, China;
2. School of Information Science and Engineering, Central South University, Changsha 410083, China
 全文: PDF  HTML
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%.
Abstract:

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%.
出版日期: 2010-07-01
:  TM 401  
基金资助:

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） .
作者简介: 胡志坤（1976—）,男,湖北鄂州人,副教授,从事复杂动态系统故障诊断技术研究.E-mail:huzk@mail.csu.edu.cnDOI: 10.3785/j.issn.1008973X.2010.07.017
服务　　
把本文推荐给朋友
加入引用管理器
E-mail Alert
RSS
作者相关文章　　

引用本文:

胡志坤, 王美铃, 桂卫华, 阳春华, 丁家峰. 基于支持向量机的时序周波波形分类方法[J]. J4, 2010, 44(7): 1327-1332.

HU Zhi-Kun, WANG Mei-Ling, GUI Wei-Hua, YANG Chun-Hua, DING Jia-Feng. Support vector machine based classification method for
timeseries periodic waveform. J4, 2010, 44(7): 1327-1332.

链接本文:

http://www.zjujournals.com/eng/CN/10.3785/j.issn.1008-973X.2010.07.017    或    http://www.zjujournals.com/eng/CN/Y2010/V44/I7/1327

［1］ LIU Y H. Wavelet feature extraction for highdimensional microarray data ［J］. Neurocomputing, 2009, 72(4/6): 985990.
［2］ LIAN X, CHEN L. Efficient similarity search over future stream time series ［J］. IEEE Transactions on Knowledge and Data Engineering, 2008, 20(1): 4054.
［3］ 胡志坤,徐飞,桂卫华,等.时序序列相似性度量的小波矩阵变换法［J］.控制理论与应用,2009,26(10): 11051110.
HU Zhikun, XU Fei, GUI Weihua, et al. Waveletmatrix transforming method for similarity measurement of fault waveform of electronic power devices ［J］. Control Theory and Applications, 2009, 26(10): 11051110.
［4］ HU Z K, GUI W H, YANG C H, et al. A waveletmatrix transforming method for extracting feature vectors of sequences for similarity measurement ［J］. International Journal of Control Automation, System, 2010, 8(2): 250256.
［5］ SONG L N, JI G R, CHEN J. Extraction of shell texture feature of coscinodiscus for classification based on wavelet and PCA ［C］∥ International Joint Conference on Artificial Intelligence. Piscataway, NJ, USA: IEEE, 2009: 282285.
［6］ LIN C T, HUANG C H. A complex texture classification algorithm based on gabortype filtering cellular neural networks and selforganized fuzzy inference neural networks ［C］∥ IEEE International Symposium on Circuits and Systems. Kobe, Japan: IEEE, 2005: 39423945.
［7］ MURPHEY Y L, OU G. Multiclass pattern classification using neural networks ［J］. Pattern Recognition, 2007, 40(1): 418.
［8］ 梅雪,吴为麟.基于小波和ANN的电能质量分类方法［J］.浙江大学学报：工学版,2004,38(10): 13841386.
MEI Xue, WU Weilin. Power quality classification based on wavelet and artificial neural network ［J］. Journal of Zhejiang University: Engineering Science, 2004, 38(10): 13841386.
［9］ GAO G H, ZHANG Y Z, ZHU Y, et al. Hybrid support vector machinesbased multifault classification ［J］. Journal of China University of Mining and Technology, 2007, 17(2): 246250.
［10］ 杜树新,吴铁军.模式识别中的支持向量机方法［J］.浙江大学学报：工学版,2003,37(5): 522527.
DU Shuxin, WU Tiejun. Support vector machines for pattern recognition ［J］. Journal of Zhejiang University: Engineering Science, 2003, 37(5): 522527.
［11］ GU H, WANG H W. Fuzzy prediction of chaotic time series based on singular value decomposition ［J］. Applied Mathematics and Computation, 2007, 185(2): 11711185.
［12］ LIU Y H, YOU Z S, CAO L P. A novel and quick SVMbased multiclass classifier ［J］. Pattern Recognition, 2006, 39(11): 22582264.
［13］ SUNGMOON C, SANG H O. SVM with binary tree architecture for multiclass classification ［J］. Neural Information Processing Letters and Reviews, 2004, 2(3): 4751.
［14］ EVGENIOU T, PONTIL M, ELISSEEFF A. Leave one out error, stability, and generalization of voting combinations of classifiers ［J］. Machine Learning, 2004, 55(1): 7197.
No related articles found!