Adaptive Gaussian Process Regression Based Remaining Useful Life Prediction of PEMFC Incorporating An Improved Health Indicator

Abstract

Considering the importance of the proton exchange membrane fuel cells(PEMFC) to daily life and industry, this paper makes the remaining useful life(RUL) prediction of the PEMFC based on two different environments. To this end, the improved health indicator is proposed to describe the health state of PEMFC. On this basis, a data-driven method, namely the adaptive Gaussian process regression(GPR) method, is proposed to predict the RUL of PEMFC. The effectiveness of the proposed life prediction method is demonstrated in the aging data set of PEMFC provided by the prognostic and health management(PHM) challenge by a case study, the artificial neural network(ANN) method, and the adaptive GPR method are used to predict the PEMFC's RUL. Results show that the adaptive GPR method achieves better prediction results and provides the probability distribution of the results compared with the ANN method.