Modeling of CO influence in PBI electrolyte PEM fuel cells

In most PEM fuel cell MEA's Nafion is used as electrolyte material due to its excellent proton conductivity at low temperatures. However, Nafion needs to be fully hydrated in order to conduct protons. This means that the cell temperature cannot surpass the boiling temperature of water and further this poses great challenges regarding water management in the cells. When operating fuel cell stacks on reformate gas, carbon monoxide (CO) content in the gas is unavoidable. The highest tolerable amount of CO is between 50-100 ppm with CO-tolerant catalysts. To achieve such low CO-concentration, extensive gas purification is necessary; typically shift reactors and preferential oxidation. The surface adsorption and desorption is strongly dependent upon the cell temperature. Higher temperature operation favors the CO-desorption and increases cell performance due to faster kinetics. High temperature polymer electrolyte fuel cells with PBI polymer electrolytes rather than Nafion can be operated at temperatures between 120-200 degrees C. At such conditions, several percent CO in the gas is tolerable depending on the cell temperature. System complexity in the case of reformate operation is greatly reduced increasing the overall system performance since shift reactors and preferential oxidation can be left out. PBI-based MEA's have proven long durability. The manufacturer PEMEAS have verified lifetimes above 25,000 hours. They are thus serious contenders to Nafion based fuel cell MEA's. This paper provides a novel experimentally verified model of the CO sorption processes in PEM fuel cells with PBI membranes. The model uses a mechanistic approach to characterize the CO adsorption and desorption kinetics. A simplified model, describing cathode overpotential, was included to model the overall cell potential. Experimental tests were performed with CO-levels ranging from 0.1% to 10% and temperatures from 160-200 degrees C. Both pure hydrogen as well as a reformate gas models were derived and the modeling results are in excellent agreement with the experiments.