Adaptive state of health estimation for lithium-ion batteries using impedance-based timescale information and ensemble learning

Abstract

Accurate state of health (SOH) estimation is vital to ensure safe, reliable, and efficient operation of lithium-ion batteries. Timescale information on internal kinetic processes is closely related to battery health and can be efficiently identified from broadband impedance by the distribution of relaxation times (DRT). Despite its bright prospects for onboard scenarios, the application of impedance-based timescale information in SOH estimation has rarely been explored. Motivated by this, this work proposes a novel SOH estimation method utilizing impedance-based timescale information and ensemble learning. Adaptive SOH estimation over a wide state of charge (SOC) range is achieved without knowledge of usage history or SOC. Specifically, timescale identification of 584 impedance spectra is conducted via DRT. Based on the extracted timescale features, an ensemble learning technique with a base learner of the regression tree is employed to estimate SOH. The average of the mean absolute errors at different SOCs can be within 1.87 % for all cells under different cyclic conditions. The minimum redundancy maximum relevance algorithm assists in the comparative investigation into SOH estimation using different feature combinations. This work exhibits excellent applications of impedance-based timescale information in SOH estimation and can provide a fresh viewpoint in promoting the health management development.