Gaussian Processes for Fast and Accurate Measurements of the Polarization Resistance of Hydrogen Fuel Cells from Impedance Spectroscopy

Abstract Polarization resistance serves as a critical performance indicator for electrochemical devices, including proton exchange membrane fuel cells. Conventionally determined from electrochemical impedance spectroscopy as the diameter of the impedance spectrum in the Nyquist representation, accurate estimation remains challenging, especially when the measured impedance is noisy, or its spectrum truncated. Herein, we present a Gaussian process framework based on the distribution of relaxation times (DRT) that leverages experimental data for robust estimation of polarization resistance with quantified uncertainty. By determining both the lowest necessary frequency and optimal frequency sampling, we established a framework for obtaining quick and reliable measurements of this resistance. This approach advances the rapid characterization of fuel cells through probabilistic analysis of impedance spectra.