基于PCA-SVM模型的多电平逆变系统故障诊断  

Fault diagnosis in multilevel inverter system based on the model of the PCA-SVM

作　　者：刘远[1] 王天真[1,2] 汤天浩[1] 李继方[1,3] 陈嘉琦[1] 

机构地区：[1]上海海事大学电气自动化系,上海200135 [2]法国海军学院研究所SIG课题组,布雷斯特29240 [3]开封大学机电工程学院,河南开封475000

出　　处：《电力系统保护与控制》2013年第3期66-72,共7页Power System Protection and Control

基　　金：国家自然科学基金(61203089);交通部资助项目(2009-329-810-030);上海市自然科学基金资助项目(11ZR1415100);上海市教育委员会重点学科建设项目(J50602)~~

摘　　要：多电平逆变系统由许多开关器件组成,具有非线性因素,很难建立数学模型来识别故障。针对多电平逆变系统故障的特点,为了实现级联H-桥多电平逆变系统的故障诊断,提高诊断的精确度和运算速度,提出了一种基于PCA-SVM模型的故障诊断策略。分析了故障信号,采用快速傅里叶变换(FFT)实现对故障信号的预处理,主元分析(PCA)提取起主要作用的特征向量,降低噪声,减少训练和测试时间,通过SVM进行训练和测试。实验结果表明,该方法具有很好的故障诊断能力、整体的诊断精度高达98%、计算效率高,满足了级联H-桥多电平系统诊断故障要求。It is difficult to diagnose fault using a mathematical model because MLIS consists of many switch devices and their complexity has a nonlinear factor. For the characteristics of faults of MLIS, in order to realize the cascade H-bridge multilevel inverter system （MLIS） fault diagnosis and improve the accuracy of the fault diagnosis and operation speed, a fault diagnosis strategy based on the model of the PCA-SVM is proposed. This paper firstly analyzes the fault signal and uses the Fast Fourier transform （FFT） technology to realize the preprocessing of the fault signal. Then the main feature vectors are extracted to reduce the influence of the noise, and decrease the training and test time. Finally, it is trained and tested by the model of PCA-SVM. Experimental results show that the diagnosis method has a very good ability and efficiency, and the overall diagnosis accuracy is more than 98%. It meets the cascade H-bridge multilevel inverter system requirements.

关 键 词：级联H-桥 多电平逆变 主元分析 支持矢量机 故障诊断 

分 类 号：TM464[电气工程—电器]