Electrochemical impedance spectroscopy (EIS) provides a detailed non-invasive insight into reaction and transport processes inside a proton exchange membrane fuel cells (PEMFCs). Measurements at different frequencies can provide information on electrochemical reactions and charge and mass transport occurring on different timescales. Combined with a physics-based model, these data can help distinguish the underlying processes and estimate selected model parameters. The accuracy of parameter estimation, therefore, depends more on which frequencies are measured than on the total number of points.
Impedance spectra are usually measured using a predefined logarithmic frequency grid. Such measurements may contain many similar or weakly informative points, while the low-frequency part considerably extends the measurement time. Specific parameters of the models, which are used to decipher these EIS spectra, may therefore remain poorly identifiable despite the large dataset and time and resources invested in its procurement, which results in less accurate analyses of intra-FC phenomena or even in inability to identify specific phenomena.
In the paper “Advancing EIS-based characterization of PEMFCs through model-based DoE and parameter sensitivity analysis”, published in the Journal of Power Sources (IF=8.4), the researchers of the University of Ljubljana, Faculty of Mechanical Engineering have developed a model-based method for designing PEMFC impedance measurement campaign. The method relies on utilization of model-based DoE and parameter analysis leveraging physics-based analytical impedance model and Fisher information analysis. This unique approach enables tailoring the frequency spectrum to the selected model and diagnostic objective instead of using the same number of points in every measurement.
The analysis showed that methodologically driven selection of highly informative frequencies can reduce the predicted variance of estimated parameters by more than an order of magnitude (in a specific study by factor of 28) while reducing measurement time by approximately a factor of 100. This as a result turns EIS test planning from an experience-driven procedure or procedure addressed with brute force approach into a transparent quantitative optimization problem that supports faster PEMFC characterization and more reliable parameter identification for advanced diagnostics and State-of-X estimation.
