Short-term performance degradation prediction of proton exchange membrane fuel cell based on discrete wavelet transform and gaussian process regression

Proton exchange membrane fuel cell (PEMFC) is regarded as one of the most promising energy conversion devices, but cost and durability are two challenges that hinder its large-scale application. Short-term performance degradation prediction is of great importance for the optimization of operating conditions and maintenance strategies to improve the durability and reduce costs of the PEMFCs. This paper proposes a novel data-driven short-term prediction method combining discrete wavelet transform (DWT), gaussian process regression (GPR) and particle swarm optimization (PSO) algorithm. The DWT decomposes the voltage time series waveform into numerous sub-waveforms with different characteristics. The GPR is utilized to construct individual prediction models for distinct sub-waveforms and the final prediction outcomes are obtained by adding the results of each model. The PSO achieves the automatic hyper-parameters optimization in order to improve the accuracy and robustness of the GPR model. In addition, the effectiveness of the proposed method is fully validated by the degradation datasets of PEMFCs in constant and quasi-dynamic load current conditions as well as real road working conditions. Finally, compared with widely used data-driven models such as artificial neural networks (ANN), support vector machine (SVM), long short-term memory (LSTM) and newly proposed methods in relevant studies, the proposed method exhibits superior accuracy and stability for degradation prediction.