A data-fusion-model method for state of health estimation of Li-ion battery packs based on partial charging curve

Abstract

The estimation of State of Health (SOH) for battery packs used in Electric Vehicles (EVs) is a complex task with significant importance, accompanied by several challenges. This study introduces a data-fusion model approach to estimate the SOH of battery packs. The approach utilizes dual Gaussian Process Regressions (GPRs) to construct a data-driven and non-parametric aging model based on charging-based Aging Features (AFs). To enhance the accuracy of the aging model, a noise model is established to replace the random noise. Subsequently, the state-space representation of the aging model is incorporated. Additionally, the Particle Filter (PF) is introduced to track the unknown state in the aging model, thereby developing the data-fusion-model for SOH estimation. The performance of the proposed method is validated through aging experiments conducted on battery packs. The simulation results demonstrate that the data-fusion model approach achieves accurate SOH estimation, with maximum errors less than 1.5%. Compared to conventional techniques such as GPR and Support Vector Regression (SVR), the proposed method exhibits higher estimation accuracy and robustness.