Loss mechanisms in PEM Fuel Cells related to charge transfer reactions or diffusive gas transport result in a strongly nonlinear current/voltage-relation, which is furthermore affected by nonlinear dependencies on operating parameters as temperature, relative humidity of the gases and stoichiometries. These dependencies have to be considered and validated in fuel cell models.

Direct comparison of simulated and measured current/voltage-relation only allows to evaluate deviations in resulting cell voltages but exclude internal state variables of the model such as overvoltages due to different loss mechanisms. In consequence, simulated and measured values of voltage or current can concur by unnoticed compensation of different errors in magnitude and sign. Additionally, limited numbers of measurements and comparisons increase the probability and impact of this effect.

Aggravating, every single process within the cell depends on the comprehensive combination of all operating conditions and runs simultaneously with all other ones. A direct comparison of individual simulated and measured polarization curves therefore is not a suitable and sufficient validation.

We address this challenge by applying a statistical methodology using design of experiments to deconvolute the significant dependencies for cell performance on the respective operating conditions. To this end, experiments are conducted with systematically and purposefully multidimensional varied operating conditions over the entire operating range. [1] Using this independent experimental approach, the impact of changing one or more operating conditions simultaneously on cell performance are determined, resulting in a comprehensive database.

This allows a close-meshed validation of the sensitivity of each operating condition to the respective dependencies represented in the model to be validated.

Furthermore, a more detailed progression of the measured data-based cell voltage in comparison to simulated internal state variables such as loss-induced overvoltages is revealed and the weighting of the modeled processes is verified.

In this contribution the application of this methodology on an impedance-based cell model presented in [2] is shown. The resulting physical interpretation is discussed, consequences and conclusions for the further development of the cell model are demonstrated.

[1] P. Oppek et al., Proc. 8th Eur. PEFC & Electrolyzer Forum, p. A1405 (2021).

[2] T. Goosmann et al., Proc. 8th Eur. PEFC & Electrolyzer Forum, p. A1307 (2021).