Dynamic thermal model development of direct methanol fuel cell

Abstract

Direct methanol fuel cell (DMFC) is fueled with liquid methanol coupled with air to produce power at reasonably lower operational conditions while resulting in by-products of carbon dioxide and water, which is more environmentally friendly. Due to the complexity associated with the performance of direct methanol fuel cell, the application of artificial neural network (ANN) can significantly predict the characteristic performance of the cells. Nevertheless, limited studies have delved into the exploration of artificial neural network in the prediction of the transient characteristics of direct methanol fuel cells. The current study however presents a detailed investigation into the prediction of the dynamic thermal characteristics of a direct methanol fuel cell stack subjected to varying operational environment. Parameters considered in the study as input include methanol concentration, anode as well as cathode inlet flow rates, coupled with current. Outcomes for the artificial neural network models for three varying learning algorithms were ascertained for anode and cathode temperatures, which were forecasted closely by models with higher number of hidden neurons. Such models have coefficients of determination of 0.95 or more and mean square error less than 0.04. Thus, the outcome of the study presents prospects for artificial neural network methods as optimum control approach in direct methanol fuel cell development.