基于数据驱动和改进光谱算法的质子交换膜燃料电池堆功率密度优化

Abstract As a green power conversion device, the power performance of proton exchange membrane fuel cell (PEMFC) stack is determined by the actual operating parameters. The optimization of the power density and corresponding operating parameters of the PEMFC according to the target demand is essential. In this paper, a global optimization strategy for the power density of PEMFC stack is proposed, which combines the random forest algorithm (RF) and the improved light spectrum optimization algorithm (ILSO). A dataset is constructed based on the simulation results of the PEMFC mathematical model and used to train a data-driven surrogate model. The input variables of the surrogate model are identified, including operating temperature, anode pressure, cathode/anode relative humidity and current density, and the output is power density. Prediction performance shows that the mean absolute error (MAE), mean square error (MSE), and coefficient of determination (R 2 ) in the training set are 0.007, 0.000097 and 0.9991, respectively. The surrogate model has considerable accuracy compared to the original model with a relative error of 0.86 %. Additionally, the average optimization time of the surrogate model is 1716.3 s, which is reduced by 44.8 % compared to the original model. By employing this strategy, an optimal power density of 1.211 W/cm 2 is obtained and the corresponding operating parameters under various target powers are predicted.