Backpropagation Neural Network‐Based Prediction Model for Fuel Cell Thermal Management System

Abstract

The temperature prediction of fuel cell thermal management systems (FCTMS) has been the focus of research in fuel cell vehicle. Herein, a prediction model of FCTMS based on the backpropagation neural network is proposed. Predictive models are applied to fuel cell TMSs for predicting temperature changes within the system. First, the fuel cell TMS is established based on Amesim software and verified by using experimental data. Then a prediction model is established based on the simulation data of the system model. After the validation calculation, the highest accuracy of the stack temperature prediction was found, with a relative error of 0.75%. The heat sink outlet temperature prediction accuracy is the worst, with a relative error of 4.3%. The mean square error of the overall output of the prediction model is 0.043, and the mean absolute percentage error of the three results is 0.23%, 0.48%, and 0.16%. Both are below 5%. Therefore, the prediction model has more precise prediction performance, which helps the parameter study and control decision setting of FCTMS.