Online generation of full-frequency electrochemical impedance spectra for Lithium-ion batteries using early-stage partial relaxation voltage curve

Electrochemical impedance spectroscopy (EIS) serves as a powerful non-destructive tool for lithium-ion battery state assessment, yet its real-time application faces significant challenges including expensive hardware requirements, time-consuming measurements, and stringent data quality demands. This study develops a hardware-free online electrochemical impedance spectroscopy using only relaxation voltage, achieved through a physics-informed neural network (PINN) that predicts full-frequency EIS from early-stage partial relaxation curves. The proposed approach exhibits remarkable insensitivity to battery state of charge and state of health, as validated by a comprehensive dataset containing over 300 impedance spectra from four batteries under various aging conditions. Experimental results demonstrate accurate EIS prediction with relative errors (RE) below 5.6 % and mean absolute errors (MAE) below 1.12 mΩ when using complete relaxation curves. Crucially, the method maintains reliability under practical constraints, achieving maximum RE of 6.1 % and MAE of 1.29 mΩ even with limited sampling data and shortened relaxation curves. By enabling online full-frequency EIS acquisition through relaxation voltage signals without hardware requirements, this work establishes a new paradigm for real-time battery diagnostics, providing valuable insights for state estimation and fault detection in battery management systems.