Distortion-aware physics-based impedance modeling of commercial Li-ion batteries under non-ideal excitation signals

Electrochemical impedance spectroscopy (EIS) provides a sensitive probe for investigating internal physicochemical processes in lithium-ion batteries and has been widely considered for diagnostic applications. However, impedance measurements obtained under practical conditions are often affected by non-ideal excitation signals, environmental noise, and systematic distortions, which complicate the interpretation of EIS data using physics-based electrochemical models. In this study, a distortion-aware framework is developed for physics-based impedance modeling of commercial 21,700 lithium-ion batteries. A pseudo-two-dimensional (P2D) model with double-layer capacitance is employed to simulate impedance responses over a broad frequency range. A frequency-resolved parameter sensitivity analysis is conducted to quantify the contributions of key electrochemical parameters across the impedance spectrum and to identify the dominant factors governing different frequency regimes. Building on this sensitivity analysis, the model is further evaluated under realistic measurement conditions. Ambient noise and waveform distortions are identified through hardware impedance testing, reconstructed as biased excitation inputs, and processed using an FFT-based approach to extract impedance spectra from time-domain signals. The model shows excellent agreement with experiments in both the DC and AC validation sets. Under low-rate discharge conditions (0.1C), the simulated voltage profile matches the experimental response, with an RMSE of 0.0423 V, and most deviations occur near the end-of-discharge knee at very low SoC. For impedance validation, the model reproduces the measured spectra with RMSE values of 0.282 mΩ and 0.224 mΩ for the real and imaginary parts, respectively. Under distorted excitations, including non-sinusoidal waveforms, Gaussian noise, and clipping, the median complex least-squares (CLS) error stays stable between 0.330 and 0.367 mΩ, and the maximum CLS error remains below 0.49 mΩ. This work provides a practical methodology for interpreting impedance spectra of commercial lithium-ion batteries when signal distortions and noise cannot be neglected.