A Bayesian transfer learning framework for assessing health status of Lithium-ion batteries considering individual battery operating states

The rapid and accurate assessment of the state of health (SOH) of lithium-ion batteries (LIBs) is of key importance for efficient battery monitoring and management. The degradation of LIBs is a complex process, with each LIB exhibiting a unique degradation path influenced by a combination of internal and external factors. However, the existing methods typically treat each LIB as a standalone entity, failing to fully leverage the unique characteristics of each individual. To address this limitation, a Bayesian transfer learning framework is proposed in this study to model the distinct LIB degradation process in order to complete the assessment of SOH. Specifically, a mixed-effect model (MEM) is constructed to describe the process of degradation of LIB health states, where the heterogeneity among each LIB can be captured. The B-spline basis functions are occupied to map the relationship between the operational factors and mixed effects. Subsequently, a covariance matrix is modeled to realize information transmission by the similarity between each LIB. To adapt to different practical application scenarios, three parameter updating strategies based on Bayes' theorem are proposed for LIB SOH estimation under practical applications with various available information. Furthermore, two datasets are used to illustrate the method's versatility, as it is applicable to diverse battery types and to both cycle and calendar ageing.