JIANG Daiyan, JIN Yuhong, ZHANG Ziheng, LIU Jingbing, ZHANG Yuan, LI Siquan, WANG Hao
The cascading utilization of retired power lithium batteries (with a rated capacity of over 80%) can effectively alleviate the pressure of battery recycling and environmental pollution, and improve resource utilization efficiency and economic benefits. However, conducting rapid, non-destructive, and accurate state assessment of the retired batteries remains a challenge. Compared with other reported methods, electrochemical alternating current measurement of batteries and collecting data to draw impedance spectra are the core methods for studying battery states, which have two advantages: fast and non-destructive. The battery detected in this way can establish internal impedance and state correlation, and quickly complete battery state evaluation. The analysis methods of electrochemical impedance spectroscopy mainly include predicting impedance based on measurement data and machine learning methods, analyzing the changes in various equivalent components of the circuit based on equivalent circuit diagrams, and using integration algorithms to convert impedance spectroscopy into a more intuitive relaxation time distribution spectroscopy. These methods all provide analytical methods for the internal aging of batteries, providing an electrochemical basis for the relationship between the internal impedance and health status of batteries. Based on this, this article reviewed the latest research progress in combining electrochemical impedance spectroscopy with machine learning to evaluate the state of power lithium batteries both domestically and internationally, with a focus on summarizing and exploring the relationship between electrochemical impedance spectroscopy, equivalent circuit models, relaxation time distribution, and machine learning.
