Enhancing model characterization of PEM Fuel cells with human memory optimizer including sensitivity and uncertainty analysis

This paper presents a novel attempt to identify the seven unknown proton exchange membrane (PEM) Fuel Cells (PEMFCs)’ parameters. The sum of quadratic deviations (SQD) between the appropriate estimated model-based and the measured dataset points is used to define the cost function. A human memory optimizer (HMO) is employed to decide on the best PEMFC parameters within acceptable boundaries. The AVISTA SR-12, BCS 500-W, NedStack PS6 6-kW, and 250-W units are four different real-world datasets of commercial PEMFCs stacks that are used to test the applied HMO method. The SQD’s values for AVISTA SR-12, BCS 500-W, NedStack PS6 6-kW and 250-W units are 0.000142335, 0.0116978, 2.145700, and 0.331371, respectively (all in $V^2$). The findings demonstrate that the PEMFC model is accurately characterized by the HMO, with sensitivity analysis performed using Monte-Carlo indicators, Sobol indices, and sensitivity metrics. The HMO-based approach has good efficacy in obtaining smooth convergence patterns and the lowest values of SQDs.