基于支持向量回归的VPMCD方法及其在局部放电模式识别中的应用
|
摘要
针对传统VPMCD方法在回归预测过程中存在的缺陷,采用支持向量回归代替原方法中的多项式回归模型,解决了原回归方法对高维小样本数据预测精度差的问题。在基于支持向量回归的VPMCD方法中,首先采集各个不同类型的放电样本,并提取特征向量构成样本集合;其次,通过支持向量回归对训练样本进行训练,建立各放电类型的变量预测模型;然后,利用这些模型对测试样本进行回归预测,得到各样本相应的预测平方和误差;最后,以预测误差平方和最小为依据,识别各放电样本的放电类型。相较于BP神经网络、SVM、传统VPMCD方法,SVR-VPMCD方法具有更好的分类效果。
Aiming at the shortcomings of the traditional Variable Predictive Model Based Class Discriminate( VPMCD) method in regression prediction,this paper substitutes the Support Vector Regression( SVR) for the original polynomial regression model. The new alternative solves the low-prediction-accuracy problem for high-dimensional small sample data in the original regression method. There are four steps in SVR-VPMCD method. The first step is to collect various discharge samples and extract feature vectors to constitute sample sets. Next step is to train samples by SVR method and construct Variable Predictive Models( VPMs) for each types. The third step is to conduct regression prediction on testing samples by these VPMs and obtain the corresponding forecasting error sum of square. The final step is to tell different types of partial discharge samples based on the minimal forecasting errors. Compared with Back Propagation( BP) neural network,Support Vector Machine( SVM),and traditional VPMCD method,SVR-VPMCD method produces better classification.
Aiming at the shortcomings of the traditional Variable Predictive Model Based Class Discriminate( VPMCD) method in regression prediction,this paper substitutes the Support Vector Regression( SVR) for the original polynomial regression model. The new alternative solves the low-prediction-accuracy problem for high-dimensional small sample data in the original regression method. There are four steps in SVR-VPMCD method. The first step is to collect various discharge samples and extract feature vectors to constitute sample sets. Next step is to train samples by SVR method and construct Variable Predictive Models( VPMs) for each types. The third step is to conduct regression prediction on testing samples by these VPMs and obtain the corresponding forecasting error sum of square. The final step is to tell different types of partial discharge samples based on the minimal forecasting errors. Compared with Back Propagation( BP) neural network,Support Vector Machine( SVM),and traditional VPMCD method,SVR-VPMCD method produces better classification.
引文
[1]VENKATESH S,GOPAL S.Robust heteroscedastic probabilistic neural network for multiple source partial discharge pattern recognition-significance of outliers on classification capability[J].Expert systems with applications.2011,38(9):11501-11514.
[2]廖瑞金,袁磊,汪可,等.基于S变换和双向二维主成分分析的局部放电模式识别[J].重庆大学学报,2013,36(5):56-63.LIAO Ruijin,YUAN Lei,WANG Ke,et al.Partial discharge pattern recognition based on S transform and two-directional 2DPCA[J].Journal of Chongqing University,2013,36(5):56-63.
[3]LUO S,CHENG J.VPMCD based novelty detection method on and its application to fault identification for local characteristic-scale decomposition[J].Cluster computing,2017,20(25):1-11.
[4]姜战伟,郑近德,潘海洋,等.基于改进多尺度熵与VPMCD的滚动轴承故障诊断[J].噪声与振动控制,2017,37(3):156-161+172.JIANG Z W,ZHENG J D,PAN H Y,et al.Fault diagnosis of rolling bearings based on improved multiscale entropy and VPMCD[J].Noise&Vibration Control,2017,37(3):156-161+172.
[5]程军圣,马利,潘海洋,等.基于EEMD和改进VPMCD的滚动轴承故障诊断方法[J].湖南大学学报(自然科学版),2014,41(10):22-26.CHENG Junsheng,MA Li,PAN Haiyang,et al.Afault diagnosis method for rolling bearing based on EE-MD and improved VPMCD[J].Journal of Hunan University(Natural Sciences),2014,41(10):22-26.
[6]JOYCE A P,LEUGN S S.Use of response surface methods and path of steepest ascent to optimize ligandbinding assay sensitivity.[J].Journal of immunological methods,2013,392(1-2):12-23.
[7]罗颂荣,程军圣,杨宇.基于本征时间尺度分解和变量预测模型模式识别的机械故障诊断[J].振动与冲击,2013,32(13):43-48.LUO Songrong,CHENG Junsheng,YANG Yu.Machine fault diagnosis method using ITD and variable predictive model-based class discrimination[J].Journal of Vibration and Shock,2013,32(13):43-48.
[8]RAGHURAJ R,LAKSHMINARAYANAN S.Variable predictive models-a new multivariate classification approach for pattern recognition applications[J].Pattern recognition,2009,42(1):7-16.
[9]马利.基于时频分析和多变量预测模型的滚动轴承故障诊断方法[D].长沙:湖南大学,2015.
[10]ZHENG J D,CHENG J S,YANG Y,et al.A rolling bearing fault diagnosis method based on multi-scale fuzzy entropy and variable predictive model-based class discrimination[J].Mechanism and machine theory,2014,78.
[11]ANDREW A M.An introduction to support vector machines and other kernel‐based learning methods[J].Kybernetes,2001,32(1):1-28.
[12]VAPNIK V N.The nature of satistical learning theory.Springer-verlag New York,Inc.1999,pp.988-999.
[13]RAGHURAJ R,LAKSHMINARAYANAN S.VPMCD:variable interaction modeling approach for class discrimination in biological systems[J].FEBS letters,2007,581(5).
[14]张蒙,朱永利,张宁,等.基于变分模态分解和多尺度排列熵的变压器局部放电信号特征提取[J].华北电力大学学报(自然科学版),2016,43(6):31-37.ZHANG Meng,ZHU Yongli,ZHANG Ning,et al.Feature extraction of transformer partial discharge signals based on varitional mode decomposition and multi-scale permutation Entropy[J].Journal of North China Electric Power University,2016,43(6):31-37.
[15]赵磊,朱永利,贾亚飞,等.基于GLCM和LBP的局部放电灰度图像特征提取[J].电测与仪表,2017,54(1):77-82.ZHAO L,ZHU Yongli,JIA Yafei,et al.Feature extraction for partial discharge grayscale image based on gray level co-occurrence matrix and local binary pattern[J].Electrical Measurement&Instrumentation,2017,54(1):77-82.
[16]刘通,薛永刚,赵煦,等.变压器内单一局放PRPD谱图统计参量的主成分分析[J].南方电网技术,2014,8(5):33-37.LIU Tong,XUE Yonggang,ZHAO Xu,et al.Principal component analysis for statistical parameters of phase resolved partial discharge spectra of single partial discharge pattern in power transformers[J].Southern Power System Technology,2014,8(5):33-37.
[17]王刘旺,朱永利,贾亚飞,等.局部放电大数据的并行PRPD分析与模式识别[J].中国电机工程学报,2016,36(5):1236-1244.WANG Liuwang,ZHU Yongli,JIA Yafei,et al.Parallel phase resolved partial discharge analysis for pattern recognition on massive PD data[J].Proceedings of the CSEE,2016,36(5):1236-1244.
[2]廖瑞金,袁磊,汪可,等.基于S变换和双向二维主成分分析的局部放电模式识别[J].重庆大学学报,2013,36(5):56-63.LIAO Ruijin,YUAN Lei,WANG Ke,et al.Partial discharge pattern recognition based on S transform and two-directional 2DPCA[J].Journal of Chongqing University,2013,36(5):56-63.
[3]LUO S,CHENG J.VPMCD based novelty detection method on and its application to fault identification for local characteristic-scale decomposition[J].Cluster computing,2017,20(25):1-11.
[4]姜战伟,郑近德,潘海洋,等.基于改进多尺度熵与VPMCD的滚动轴承故障诊断[J].噪声与振动控制,2017,37(3):156-161+172.JIANG Z W,ZHENG J D,PAN H Y,et al.Fault diagnosis of rolling bearings based on improved multiscale entropy and VPMCD[J].Noise&Vibration Control,2017,37(3):156-161+172.
[5]程军圣,马利,潘海洋,等.基于EEMD和改进VPMCD的滚动轴承故障诊断方法[J].湖南大学学报(自然科学版),2014,41(10):22-26.CHENG Junsheng,MA Li,PAN Haiyang,et al.Afault diagnosis method for rolling bearing based on EE-MD and improved VPMCD[J].Journal of Hunan University(Natural Sciences),2014,41(10):22-26.
[6]JOYCE A P,LEUGN S S.Use of response surface methods and path of steepest ascent to optimize ligandbinding assay sensitivity.[J].Journal of immunological methods,2013,392(1-2):12-23.
[7]罗颂荣,程军圣,杨宇.基于本征时间尺度分解和变量预测模型模式识别的机械故障诊断[J].振动与冲击,2013,32(13):43-48.LUO Songrong,CHENG Junsheng,YANG Yu.Machine fault diagnosis method using ITD and variable predictive model-based class discrimination[J].Journal of Vibration and Shock,2013,32(13):43-48.
[8]RAGHURAJ R,LAKSHMINARAYANAN S.Variable predictive models-a new multivariate classification approach for pattern recognition applications[J].Pattern recognition,2009,42(1):7-16.
[9]马利.基于时频分析和多变量预测模型的滚动轴承故障诊断方法[D].长沙:湖南大学,2015.
[10]ZHENG J D,CHENG J S,YANG Y,et al.A rolling bearing fault diagnosis method based on multi-scale fuzzy entropy and variable predictive model-based class discrimination[J].Mechanism and machine theory,2014,78.
[11]ANDREW A M.An introduction to support vector machines and other kernel‐based learning methods[J].Kybernetes,2001,32(1):1-28.
[12]VAPNIK V N.The nature of satistical learning theory.Springer-verlag New York,Inc.1999,pp.988-999.
[13]RAGHURAJ R,LAKSHMINARAYANAN S.VPMCD:variable interaction modeling approach for class discrimination in biological systems[J].FEBS letters,2007,581(5).
[14]张蒙,朱永利,张宁,等.基于变分模态分解和多尺度排列熵的变压器局部放电信号特征提取[J].华北电力大学学报(自然科学版),2016,43(6):31-37.ZHANG Meng,ZHU Yongli,ZHANG Ning,et al.Feature extraction of transformer partial discharge signals based on varitional mode decomposition and multi-scale permutation Entropy[J].Journal of North China Electric Power University,2016,43(6):31-37.
[15]赵磊,朱永利,贾亚飞,等.基于GLCM和LBP的局部放电灰度图像特征提取[J].电测与仪表,2017,54(1):77-82.ZHAO L,ZHU Yongli,JIA Yafei,et al.Feature extraction for partial discharge grayscale image based on gray level co-occurrence matrix and local binary pattern[J].Electrical Measurement&Instrumentation,2017,54(1):77-82.
[16]刘通,薛永刚,赵煦,等.变压器内单一局放PRPD谱图统计参量的主成分分析[J].南方电网技术,2014,8(5):33-37.LIU Tong,XUE Yonggang,ZHAO Xu,et al.Principal component analysis for statistical parameters of phase resolved partial discharge spectra of single partial discharge pattern in power transformers[J].Southern Power System Technology,2014,8(5):33-37.
[17]王刘旺,朱永利,贾亚飞,等.局部放电大数据的并行PRPD分析与模式识别[J].中国电机工程学报,2016,36(5):1236-1244.WANG Liuwang,ZHU Yongli,JIA Yafei,et al.Parallel phase resolved partial discharge analysis for pattern recognition on massive PD data[J].Proceedings of the CSEE,2016,36(5):1236-1244.