Fast capacity estimation for lithium-ion battery based on online identification of low-frequency electrochemical impedance spectroscopy and Gaussian process regression

Accurate capacity estimation is critical to improving the safety and reliability of lithium-ion (Li-ion) battery systems. Traditional capacity estimation mainly extracts capacity-related features by passively collecting the battery voltage, current, and temperature signals, which requires high integrity and regularity of charging curves. In this paper, six health indicators (HIs) are extracted through the online identification of Low-frequency electrochemical impedance spectroscopy (LEIS) by step wave, combined with Gaussian process regression (GPR) to achieve a fast capacity estimation for Li-ion batteries. The step wave is injected into the battery system during the charging process through BMS and bi-directional converter cooperation. Compared with square and multi-sine waves, the stress and response of each frequency step wave are equivalent to that of the sine wave. Moreover, HIs are resolved from the actual Warburg impedance angle instead of the empirical angle π/4. Three novel HIs related to the Li-ion diffusion coefficient are proposed: Warburg factor W_d, pseudo-Li-ion diffusion state PLDS and residual signal of empirical mode decomposition PLDSr. The volume of GPR training data is only 34% of the whole frequency band EIS data. The results show that the identified LEIS achieves 0.96 goodness-of-fit (R²) at the minimum sampling frequency of 50 Hz, and the novel HIs significantly improve the state of health (SOH) estimation accuracy with R² above 0.95, root mean squared error below 1%, and mean absolute percentage error of about 0.9%. This method is an effective way to the active SOH detection for Li-ion battery, which is vital for the online SOH evaluation and early safety warning.