In this dissertation a direct methanol fuel cell stack model is described that resolves the electrochemical performance in individual cells, while having the capability to capture large scale effects across an entire stack through the use of a supercomputer. Typically stack sizes of 10 to 100 cells are modelled by dividing up cells into elements that are then distributed amongst the central processing units of the available cluster. The model contains thermal generation, polarisation losses in the bipolar plate, and gas expansion in the anode stream, modelled as a continuum.Using this model a number of studies were undertaken: A parametric study looking at optimising reactant feed in which the methanol stoichiometry and concentration were varied, which also served also as a method of validation for the model against test data collected in parallel. The second study looks at fuel starvation, and what happens when one half of a cell in a stack experiences a reduction in feed stoichiometry and the impact across neighbouring cells in the stack. This highlighted critical effects that lead to smoothing of the anodic over-potential due to high in-plane conductivity of the bipolar plate, and cross-sectional current profiles that indicate bypassing of the under-feed region. The third study looks at the effect of water saturation on the cathode, and provides insight into the transition between non-critical to critical levels of water blockage in cathode gas diffusion layer, also highlighting the role that anode limiting current plays on the current density profile. The final section deals with the performance, scalability and convergence behaviour of the code, demonstrating the inherent flexibility and speed of such a method to create a holistic large-scale fuel cell model.
展开▼
