Hydrogen fuel cell diagnostics using random forest and enhanced feature selection

To know the health status of hydrogen fuel cell, some diagnostics methods are proposed based on the historical status data of the hydrogen fuel cell. As multiple factors would cause the fuel cell problem, feature selection would be necessary during the diagnostics. In this paper, the author tried to apply an enhanced PCA algorithm to generate the proper features. Based on these features, a random forest algorithm is constructed for predicting the health status based on the history data. In this paper, we explore all aspects of hydrogen fuel cell sensor data, and extract several features by performing statistical analysis. We propose an efficient and accurate model for hydrogen fuel cell diagnostics. This model supports pipeline operations from feature selection to final result prediction. Besides, the model can also show which factor is essential for the health status of hydrogen fuel cell. (C) 2019 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.