Model prediction of dynamic performance response of DMFC using artificial neural networks

Abstract

Direct methanol fuel cell (DMFC) uses liquid methanol as fuel to generate electricity at low operating temperatures as well as to mainly produce water and carbon dioxide. Since DMFC performance characteristics are inherently complex, it can be postulated that artificial neural networks (ANN) represent a marked improvement in prediction capabilities. However, very little investigation has been done to develop dynamic ANN to predict transient behavior of DMFCs. This paper predicts the dynamic performance of a DMFC stack under changes in operating conditions including step change in current. Input variables for the analysis consist of methanol concentration, temperature and current. The performances of the ANN models of four different approaches are judged based on stack voltage, which was shown to be predicted. The results show promise of ANN modeling approaches for optimal control strategy development in DMFC system applications.