Prediction of PEMFC stack aging based on Relevance Vector Machine

Proton Exchange Membrane Fuel Cell (PEMFC) systems have been proved to be promising energy sources representing other conventional energy sources. However the life span is limited to some factors like intolerance of impurities or oscillation of working conditions, which can lead to output voltage ageing over operation. The prediction of output voltage drop trends is one of the major tasks of PEMFC system management. In this work, a prediction method of Relevance Vector Machine (RVM) is proposed, which can either give good accuracy and a confidential interval. Firstly the mathematical theory is explained thoroughly, and then the RVM is implemented to predict two voltage dropping trends based on two degradation data of a PEMFC. Finally the results are discussed and the effectiveness is evaluated. The RVM is proved to be a good candidate to predict the degradation trends of PEMFC.