基于TCN-GRU-A模型的PEMFC老化预测方法

Degradation prediction method for PEMFC based on the TCN-GRU-Amodel

为实现对动态工况下质子交换膜燃料电池(PEMFC)的老化预测,并提升门控循环单元网络(GRU)的预测能力,提出了一种将时间卷积网络(TCN)、注意力机制(Attention)和GRU结合起来的TCN-GRU-A预测方法,通过引入TCN层提升GRU特征提取能力,并用注意力机制对GRU输出特征进行加权以提升预测精度。采用PEMFC动态耐久实验数据集进行验证,通过与多种深度学习模型的预测结果对比表明:在对全电流负载数据和定电流负载数据进行的预测中,该预测方法均具有更小的预测误差和更好的拟合度。

In order to predict the aging of proton exchange membrane fuel cells(PEMFCs)under dynamic operating conditions and improve the prediction ability of the gated recurrent unit network(GRU),this paper proposes a TCN-GRU-A prediction method that combines time convolutional network(TCN),attention mechanism,and GRU.By introducing the TCN layer to enhance the feature extraction ability of GRU,the attention mechanism is used to weight the output features of GRU to improve the accuracy of the prediction.Validated using a PEMFC dynamic durability test dataset,a comparison with various deep learning models'predictions indicates that the proposed method demonstrates lower prediction errors and better fitting,whether applied to full-current load data or constant-current load data.