Rapid Estimation of Lithium-Ion Battery State of Health Based on Gaussian Process Dynamic Models

The State of Health (SOH) of lithium-ion batteries is crucial for the safe and reliable operation of electric vehicles. However, most existing SOH estimation methods require extensive and complete battery aging data. Particularly in the early stages of a battery's lifespan, the lack of data covering the entire lifecycle presents a significant challenge for accurate SOH estimation. In this study, we develop a rapid and accurate method to estimate SOH using an improved Gaussian Process Dynamic Model (GPDM) to overcome this limitation. First, latent health features that are highly correlated with the battery's capacity over its entire lifecycle are extracted from the charging data. Next, we introduce a kernelized version of GPDM to provide a more expressive covariance structure between observable co-ordinates and latent coordinates. This allows the integration of different temporal features into distinct physical observables without requiring values beyond the available data timeframe. By combining these health features with GPDM, we can track the future health status of the battery until the end of its life. Finally, the proposed algorithm is validated using NASA's battery dataset. The results demonstrate that this method can achieve accurate SOH estimation over the entire lifecycle of the battery using only a small amount of early-life data. This confirms the effectiveness and reliability of our approach for early-stage SOH estimation.