Prediction of PEMFC Life Based on TCN-BiGRU-Attention

To enhance the accuracy of the life prediction of the Proton Exchange Membrane Fuel Cell (PEMFC), This paper proposes a life prediction method based on Temporal Convolutional Neural Network (TCN), Bidirectional Gated Recurrent Unit (BiGRU), and Self-Attention mechanism (Attention). Firstly, the method fully captures the temporal features in the PEMFC data sequence through TCN. Secondly, the method utilizes BiGRU to deeply explore the complex dependencies within the data sequence. Finally, the self-attention mechanism is introduced to focus on key data. The effectiveness and accuracy of the proposed method in this paper were verified through simulation. The simulation results show that when the training set accounts for 60%, compared with the TCN model, the root mean square error and mean absolute error of the predicted voltage of the proposed method are reduced by 33.3% and 34.8% respectively, and the deviation between the predicted life value and the actual value is smaller, which improves the accuracy of life prediction.