Hybrid stochastic-deterministic algorithms for the interpretation of Electrochemical Impedance Spectroscopy spectra of Proton Exchange Membrane Fuel Cells

This work presents a case study on the use of three hybrid algorithms combining a stochastic part -Genetic Algorithms (GA), Particle Swarm Optimization (PSO), or Simulated Annealing (SA)- with a deterministic Nelder-Mead (NM) algorithm for the estimation of the equivalent electrical circuit (EEC) parameters for the interpretation of Proton Exchange Membrane Fuel Cell (PEMFC) impedance data. These hybrid methods were evaluated on mathematical test functions as well as for the interpretation of simulated and experimental PEMFC impedance spectra using EEC of different complexity.

The three stochastic/deterministic methods were compared in terms of stability, efficiency, ability to explore multiple solutions, and computing resources. The results showed that all hybrid methods were able to improve the interpretation of experimental EIS data by identifying satisfying and physically meaningful solutions, with low least-square residuals and by reducing the sensitivity to initial conditions while accelerating convergence. All methods allowed an improvement compared to the use of one single type of algorithm alone -deterministic and stochastic.