基于卷积神经网络的三电平断路故障诊断方法研究

Research on Three-level Open Circuit Fault Diagnosis Method Based on Convolutional Neural Network

逆变器作为一种电力变换的装置,具有性能优越、使用方便等优点,在生产中不可或缺。具有大功率的三电平逆变器核心元器件发生故障时,仅仅依靠人工检查很难直接判断出故障类型,存在一定的安全隐患,而且因为三电平的故障类型差异性很大,造成数据集分布不平衡的问题。针对现有数据集故障样本少、数据集不平衡的问题,本文应用合成少数类过采样技术处理数据集,并用卷积神经网络模型对三电平逆变器进行故障诊断。实验结果表明,本文使用的卷积神经网络的故障诊断准确率达96%。

As a kind of power conversion device,inverter has the advantages of superior performance and convenient use,which is indispensable in production.When the core components of high-power three-level inverter fail,it is dif⁃ficult to directly determine the fault type only by manual inspection,and there are certain security risks.Moreover,because the fault types of three-level inverter are very different,the data set distribution is unbalanced.In order to solve the problem of few fault samples and unbalanced data sets in the existing data sets,this paper applied SMOTE to deal with the data sets,and used the convolution neural network model to diagnose the fault of three-level inverter.The experimental results show that the fault diagnosis accuracy of the convolutional neural network is 96%.