首页|期刊导航|船电技术|基于二维CNN和GRU的五相逆变器故障诊断

基于二维CNN和GRU的五相逆变器故障诊断OA

Fault Diagnosis of Five-Phase Inverters Based on 2D Convolutional Neural Network-GRU

中文摘要英文摘要

针对五相电压型逆变器存在的开关器件开路故障问题,本文设计了一种具有二层卷积的CNN-GRU网络模型架构,并提出了一种基于二维卷积神经网络和门控循环单元的故障诊断方法.建立了基于正序对称分量的样本数据集,将数据集输入到二维卷积神经网络模型中,利用二维CNN-GRU单元提取数据样本的时空特征,通过输出层获得逆变器故障诊断结果.实验结果表明,该方法能有效、准确地诊断逆变器开路故障.

Aimed at the open-circuit of power switches in the five-phase voltage source inverter,the paper designs a CNN-GRU network architecture with two convolutional(2D)layers and proposes a fault diagnosis method for five-phase inverters based on 2D convolutional neural networks and gated recurrent units(GRU).A sample dataset is established based on positive-sequence symmetrical components,which is fed into the two-dimensional CNN model.The spatial-temporal features of samples are extracted by the 2D CNN-GRU units,and the fault diagnosis results of the inverter are obtained via the output layer.Experimental results show that the proposed method can diagnose open-circuit faults of the inverter effectively and accurately.

崔林涛

环球数科股份有限公司,广东 深圳 518052

信息技术与安全科学

故障诊断卷积神经网络门控循环单元五相逆变器

Fault diagnosisConvolutional neural networkGated recurrent units(GRU)Five-phase inverter

《船电技术》 2026 (9)

1-5,5

国家自然科学基金(51779102)

评论