Proton membrane fuel cell stack performance prediction through deep learning method

Proton exchange membrane fuel cell (PEMFC) system is a complex non-linear system affected by many factors with interaction, which make it difficult to build an accurate mathematical model to predict the fuel cell performance. A fuel cell performance model using a Lattice gate recurrent unit (LGRU) based on recurrent neuron network (RNN) is proposed in this paper. The proposed LGRU method can remember the previous information both in time and in depth, which help to accurately predict the performance of the fuel cell stack while reducing the computation burden. Simulation and experiment are implemented on a 2.5 kW fuel cell stack. The Root mean square error (RMSE) and of voltage prediction can reach 0.0038 and the RMSE of the temperature prediction can reach 0.0040 The fuel cell stack model established using LGRU method can be efficiently used in energy management strategy designing for accurately predicting the fuel cell stack performance.