A reliable data-driven state-of-health estimation model for lithium-ion batteries in electric vehicles

Abstract

The implementation of a precise and low-computational state-of-health (SOH) estimation algorithm for lithium-ion batteries represents a critical challenge in the practical application of electric vehicles (EVs). The complicated physicochemical property and the forceful dynamic nonlinearity of the degradation mechanism require data-driven methods to substitute mechanistic modeling approaches to evaluate the lithium-ion battery SOH. In this study, an incremental capacity analysis (ICA) and improved broad learning system (BLS) network-based SOH estimation technology for lithium-ion batteries are developed. First, the IC curves are drawn based on the voltage data of the constant current charging phase and denoised by the smoothing spline filter. Then, the Pearson correlation coefficient method is used to select the critical health indicators from the features extracted from the IC curves. Finally, the lithium-ion battery SOH is assessed by the SOH estimation model established by an optimized BLS network, where the BLS network is formed through its L2 regularization parameter and the enhancement nodes’ shrinkage scale filtrated by a particle swarm optimization algorithm. The experimental results demonstrate that the proposed method can effectively evaluate the SOH with strong robustness as well as stability to the degradation and disturbance of in-service and retired lithium-ion batteries.