基于小波包能量谱和ELM的光伏逆变器多故障在线诊断  

Online multiple fault diagnosis for PV inverter based on wavelet packet energy spectrum and extreme learning machine

作　　者：姜媛媛[1,2] 王友仁[1] 吴 孙权[1] 罗慧[3] 

机构地区：[1]南京航空航天大学自动化学院,南京210016 [2]安徽理工大学电气与信息工程学院,淮南232001 [3]南京农业大学工学院,南京210095

出　　处：《仪器仪表学报》2015年第9期2145-2152,共8页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(61371041);国家自然科学基金(61401215);江苏省普通高校研究生科研创新计划资助项目;中央高校基本科研业务费专项资金(CXLX11_0183);航空科学基金(2013ZD52055);国家商用飞机制造工程技术研究中心创新基金(SAMC14-JS-15-051)项目资助

摘　　要：以三电平光伏逆变器为研究对象,提出一种多故障模式快速诊断新方法。首先,利用小波包分解提取出三电平逆变器的桥臂电压和上、下管电压信号的能量谱特征向量,并利用主成分分析降维后获取故障特征向量;然后,基于极端学习机诊断模型分离出单器件及多器件开路等多种故障模式。实验结果表明,相比于传统BP神经网络、最小二乘支持向量机故障诊断方法,该方法检测信号易获取,抗干扰性强,诊断速度快、精度高,减小了诊断成本和复杂性,适用于在线诊断。A novel rapid diagnosis method of multiple faults for three-level PV inverters is proposed in this paper. Firstly, the wavelet packet decomposition is adopted to extract the voltage energy values of the bridge arm and upper, lower tubes in three-level PV inverter as the energy spectrum feature vectors and the principal component analysis method is used to perform the dimension reduction of the wavelet packet energy feature vectors and obtain the fault feature vectors. Then, various fault modes, such as single and multiple device open circuit and etc. are isolated based on the extreme learning machine diagnostic model. Experiment results show that compared with traditional BP neural network and least squares support vector machine fault diagnosis methods, the proposed method has the advantages of easy to obtain the detection signal, strong resistance to interference, fast diagnosis speed and high precision, and can reduce the cost and complexity of fault diagnosis. The proposed method is also suitable for online diagnosis.

关 键 词：极端学习机 小波包分解 光伏逆变器 故障诊断 特征提取 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] TH89[自动化与计算机技术—控制科学与工程]