Short-term degradation hybrid prognostics for dynamically operating fuel cells

Recently proton exchange membrane fuel cells (PEMFC) are widely used in various commercialization scenarios such as automobiles. However, there are still challenges in extracting the health indicators of PEMFC under dynamic operating conditions and then realizing accurate degradation prediction. In response to the problem, a hybrid prognostic framework is proposed in this paper. Specifically, health indicators are extracted based on a degradation mechanism model. Then, short-term degradation prediction is realized by utilizing a Bi-directional Long and Short-term Memory Network. To confirm the accuracy of the proposed method, a vehicle-oriented PEMFC durability test dataset is employed. The experimental results show that the proposed model can provide reconstructed stack voltages with up to ${9 9 \%}$ accuracy. In turn, a short-term prediction with no more than 5% relative error is realized.