| In this thesis, an Artificial Neural Networks (ANNs) system is used for the model of the Direct Methanol Fuel Cell (DMFC). This method is a novel application to this field. Currently, because of the energy crisis and environmental impacts caused by the continuous depletion of fossil fuels and the increasing exhaust gas that results from burning fossil fuels, fuel cell technologies are becoming more and more attractive. The DMFC has advantages that methanol is used directly without the methanol reforming process while the DMFC involves very complicated relations between the operational parameters and the performance of the DMFC.;In this thesis, a model of DMFC utilizing trained ANNs is presented. This model can build the forward and inverse relations between the operational parameters and the performance of the DMFC via the approximations offered by ANNs. Based on this model, several scenarios are introduced. Performance of the DMFC, including fuel cell voltage, fuel cell power density, fuel cell efficiency, and methanol crossover, is approximated by the trained ANNs in terms of methanol concentration, temperature and membrane thickness. This presents a direct relation approximation by the model. On the other hand, the parameters are successfully approximated by the trained ANNs in terms of the power density. This represents an inverse relation approximation by the model of the DMFC.;From the approximation results of the several scenarios of the DMFC model via the trained ANNs, the temperature, methanol concentration, and membrane thickness are observed to play important roles in the performance of the DMFC, indicating that these are the significant factors affecting the performance of the DMFC. Meanwhile, based on I the environmental conditions of DMFC equipment, for any power density requirement, the ANNs will give a mapping output, which will be the proper temperature or methanol concentration or membrane thickness.;Finally, the novel method enables the realization of efficient model of the direct methanol fuel cell, which can deliver a variable power demand by varying appropriately the membrane thickness, fuel concentration, and/or temperature. That is, the non-linear forward and inverse relations between the performance and the operational parameters of the DMFC model are approximated via the trained ANNs.;It is important to point out that this new ANN method, not only could be applied to the DMFC, but also could be applied to other types of fuel cells with or without explicit relationships between the power density outputs and the environmental and or operational parameter inputs. |
