A data-driven method for extracting aging features to accurately predict the battery health

Data-driven methods have been widely used for estimating the state of health (SOH) of lithium-ion batteries (LiBs). The aging process can be characterized by degrading features. To achieve high accuracy, a novel method combining four algorithms, i.e. the correlation coefficient, least absolute shrinkage and selection operator regression, neighborhood component analysis, and ReliefF algorithm, is proposed to select the most important features, which are derived from the measured and calculated parameters. To demonstrate the effectiveness of the proposed method, it is adopted to estimate the SOH of two types of LiBs: i.e. NCA and LFP batteries. Compared to the case using all features, using the selected features can improve the accuracy of SOH estimation by 63.5% and 71.1% for the NCA and LFP batteries, respectively. The method can also enable the use of data obtained in partial voltage ranges, based on which the minimum root mean square errors on SOH estimation are 1.2% and 1.6% for the studied NCA and LFP batteries, respectively. It demonstrates the capability for onboard applications.

Automated summary

In this study, a novel method combining four algorithms was proposed to select the most important features for estimating the state of health (SOH) of lithium-ion batteries (LiBs). The selected features improved the accuracy of SOH estimation by 63.5% and 71.1% for NCA and LFP batteries, respectively, compared to using all features. Additionally, the method allowed the use of data obtained in partial voltage ranges, resulting in minimum root mean square errors of 1.2% and 1.6% for NCA and LFP batteries, respectively, demonstrating its capability for onboard applications.