Efficient Parallel Simulation of Direct Methanol Fuel Cell Models

Abstract

Fuel cells, as highly efficient and environment-friendly power sources, have attracted ever-increasing research and development in the past few decades. Mathematical modeling and simulation are applied to examine the influence of the different physical and electrochemical phenomena occurring in the structure of the fuel cell during operation. Such simulation programs take a very long time (of the order of days or even weeks) to run, especially when accurate results are called for. To achieve accurate results and more insight, we need even more complex models (two- and three-dimensional models). In this paper, we present techniques for the parallel simulation of a mathematical model of direct methanol fuel cells (DMFC). In particular, we employ a paradigm called LessTalk. LessTalk is a general technique that can be employed to reduce communications in parallel computations. Using LessTalk, we have achieved almost linear speedup in simulating the DMFC model. The results were validated against measurements available from the Connecticut Global Fuel Cell Center (CGFCC).