Machine learning-based voltage degradation prediction with uncertainty quantifications for PEMFC

The voltage degradation trend prediction has been attached great importance to the application of proton exchange membrane fuel cell systems. Based on Gaussian process regression (GPR) and Bayesian neural network (BNN), this paper uses proton exchange membrane fuel cell dataset to predict the voltage degradation trend. Further, the uncertainty representation of prediction results is realized based on GPR and BNN. Finally, the mean absolute error and root mean square error are used as performance indicators to evaluate the prediction performance. The experimental results show that the evaluation indexes of BNN reach 0.00523, 0.00486 for mean absolute error and 0.007145, 0.007283 for root mean square error, which are less than that of GPR.