Multi‐step performance degradation prediction method for proton‐exchange membrane fuel cell stack using 1D convolution layer and CatBoost

Abstract

The increasing environmental issues such as climate change and air pollution require energy saving and emission reduction in various fields, such as manufacturing, building, and transportation. To address the above problem, proton‐exchange membrane fuel cells (PEMFC) gradually become promising green energy conversion device due to the advantages of zero pollution, high efficiency, and low operating noise. However, the durability problem has extremely limited the PEMFC large‐scale commercial application. To prolong the service life of PEMFC, performance degradation prediction is an effective method. This paper proposes a multi‐step performance degradation prediction method for proton‐exchange membrane fuel cells based on CatBoost feature selection, convolution computing, and interactive learning mechanism. CatBoost is used to evaluate the importance of the monitor parameters on performance degradation. The evaluation results and PEMFC degradation mechanism analyses are used to select the monitor parameters for construing the prediction model. Based on the 1D convolutional layer and the interactive learning mechanism, the prediction model is proposed to extract the deep features from the monitor data to predict the performance degradation of the fuel cell system. In particular, the multi‐step prediction is performed by the configurable sliding window. The effectiveness of the proposed method is verified on real experiment datasets, and the experiment results show that the proposed method is particularly effective for multi‐step degradation prediction and decreases the computation by feature selection and 1D convolution layer.