Online state of health estimation for Li-ion batteries in EVs through a data-fusion-model method

Accurate State of Health (SOH) estimation of batteries presents a significant challenge in ensuring the safety and durability of Electric Vehicles (EVs). Traditional methods, however, continue to face limitations due to insufficient aging models and inadequate explainability for SOH estimation. This paper introduces a novel data-fusionmodel method to address these shortcomings, specifically tailored for SOH estimation of Li-ion batteries (LIBs). To effectively capture battery degradation characteristics, two Aging Features (AFs) are extracted and analyzed, leveraging a partial charging curve during the battery's charging process. Meanwhile, a flexible and data-driven battery aging model is devised employing dual Gaussian Process Regressions (GPRs), adept at describing the nonlinear and non-Gaussian characteristics inherent in battery aging. To bolster the tracking performance of the aging model, a Rejection Sampling Particle Filter (RSPF) is proposed to mitigate the inherent fuzziness in Particle Filter (PF) measurements. This method integrates rejection sampling and PF to obtain the posterior distribution, thus enhancing filtering accuracy. Furthermore, a data-fusion-model framework is developed to amalgamate the strengths of both approaches. Experimental validation confirms the accuracy and reliability of the proposed method. Results demonstrate that the data-fusion-model method achieves high precision in estimating battery SOH, surpassing existing techniques in performance.