Risk-controlled battery parameterization from electrochemical impedance spectroscopy for model inputs and measurement planning

<jats:p>Physics-based parameterization from electrochemical impedance spectroscopy increasingly supplies parameters to battery models, yet a spectrum constrains some parameters far better than others. Across 87 spectra from 13 commercial lithium-ion cells, fits that are indistinguishable in residual disagree on diffusion quantities by 1.6 to 3.3 orders of magnitude while the ohmic resistance is essentially unchanged. A low fitting error therefore cannot license a complete parameter set, and we treat parameterization as a decision made one parameter at a time. For a 48-parameter porous-electrode impedance model, each parameter is estimated by whichever of eight regularized fits its own diagnostics favor, and is reported only when out-of-sample evidence supports it under a stated error budget. On model-generated spectra this recovers 28 of 48 parameters, replicated on three independently generated panels, and certifies about 10 parameters per run with 96 percent of them correct. Measurement design matters more than estimator strength: a balanced grid of 7 temperatures and 7 states of charge gives the certified yield of a full-frequency sweep using 36 percent of the impedance points. Applied to the 13 measured cells, the same rules return an auditable conditional parameter record and name the missing measurement for every parameter they decline.</jats:p>