经验模态分解与样本熵在并网型光伏逆变器故障诊断中的应用

Application of EMD and Sample Entropy in the Fault Diagnosis of Photovoltaic Grid Inverter

摘要:

摘要：针对光伏并网逆变器电路中故障信号的非线性、非平稳特点，提出一种基于经验模态分解(EMD)和样本熵(SampEn)的故障诊断方法。首先，利用经验模态分解对逆变器的三相输出电压进行分解，得到有限个本征模式分量(IMF)，从中选取包含故障主要信息的前几个本征模式分量提取故障信息。然后，计算本征模式分量的样本熵，从而得到用于故障诊断的特征向量；最后，将逆变器开路故障进行分类和编码，将故障特征向量输入BP神经网络进行模式识别，从而达到故障诊断的目的。在Matlab环境下对光伏并网逆变器的故障诊断进行了实验，实验结果证明了文中方法能实现对光伏并网逆变器的故障诊断，且与小波包变换相比，该方法具有诊断效率高和准确度高等特点。

Abstract:

Aiming at nonlinear and non-stationary of the fault signal of the photovoltaic grid inverter,a method of faults diagnosis was proposed based on empirical mode decomposition and sample en-tropy.Firstly, the original signal was decomposed with empirical mode decompose-tion(EMD) on the basis of theScharacteristics ofSadaptiveSmulti-resolution and a series ofSintrinsicSmode functions were obtained. The intrinsic mode functions containing the most information were chosen to extract fault information.Secondly, the sample entropy of the intrinsic mode functions containing the most information was calculated as the eigenvalues of normal signal .At last,the characteristic vectors were input BP neural network for pattern recognition to achieve the goal of fault diagnosis.The simulation results showed that the proposed method could extract fault feature effectively and the intrinsic mode functions and the sample entropy values were markedly different among different operating conditions.The feasibility and effectiveness of the method were proved by ComparedSwith theStraditionalSwaveletSpacket transform.