A feature extraction approach for state-of-health estimation of lithium-ion battery

Abstract

Accurate state-of-health (SOH) estimation is essential to ensure the reliable and safe usage of lithium-ion batteries (LIBs). A novel health feature extraction approach is proposed in this manuscript for battery SOH estimation. Firstly, the degradation data are collected from LIBs with two different life stages, and then the discrete incremental capacity (IC) curve is obtained under different constant voltage intervals dv. The corresponding charging voltage range with obvious variation trend of IC is selected and divided into several subintervals with Δv. The average IC of each subinterval is obtained. Furthermore, the consistency between the average IC of each voltage subinterval and battery capacity is analyzed and evaluated based on raw discrete IC curve. The average IC with the most consistent in relation to battery capacity degeneration is selected as the health feature. The impact of varying dv and Δv values on the feature is conducted based on Spearman correlation analysis, and the health feature with maximum Spearman correlation coefficient is used to build battery SOH estimation model. Finally, two SOH estimation models and comparative analysis of the performance between proposed health feature and other accepted features are utilized to verify the proposed health feature extraction approach. The results demonstrate that our extracted health feature effectively reveals the battery performance degeneration.