基于DCGAN的燃料电池故障诊断

Fuel cell fault diagnosis based on DCGAN

质子交换膜燃料电池(PEMFC)存在易故障、使用寿命短等缺陷。针对PEMFC常见的电堆内部水淹和膜干故障,提出基于深度卷积生成对抗网络(DCGAN)的故障诊断方法。该方法通过主成分分析(PCA),对PEMFC原始数据进行预处理,降低数据维度并提取故障特征,进而利用DCGAN对PEMFC进行正常、水淹和膜干等3种健康状态分类。该方法采用PEMFC实测数据集进行故障诊断分析,总体故障准确率为98.33%,诊断时长为2.79 s。所提出方法适用于PEMFC水管理故障诊断问题,具有一定的工程应用价值。

Proton exchange membrane fuel cell(PEMFC)had the defects such as easy failure and short service life.A fault diagnosis method based on deep convolutional generative adversarial nets(DCGAN)was proposed for the most common faults of internal flooding and membrane drying in PEMFC.The method preprocessed the original data of PEMFC through principal components analysis(PCA),reduced the data dimension and extracted fault features,then a DCGAN was used to classify the three health states of PEMFC,including normal state,flooded state and dry state.PEMFC measured data set were used for fault diagnosis analysis of this method,the overall fault accuracy was 98.33%,the diagnosis time was 2.79 s.The proposed method was suitable for PEMFC water management fault diagnosis and had certain engineering application value.