Lithium battery health state assessment based on vehicle-to-grid (V2G) real-world data and categorical boosting model

In the context of vehicle-to-grid (V2G) applications, the precise assessment of the state-of-health (SOH) of lithium-ion batteries is of paramount importance, contributing to the optimal operation of vehicles and ensuring grid stability. This paper presents a real-world SOH estimation framework utilizing the Categorical Boosting (CatBoost) algorithm. The process begins with raw data processing to extract the segments of charging and grid-feeding characterized by stable currents. Subsequently, the ampere-time integration method is employed on these segments to obtain the reference capacity, capturing the nonlinear degradation. On the basis of this, five characteristic parameters are identified as model inputs through physical significance analysis. The model's estimation performance is then compared to seven other machine learning algorithms, revealing that the proposed model offers the highest level of accuracy. It achieves the MAPE of 1.501% and RMSE of 2.279Ah, indicating its potential for effectively assessing battery health in large-scale V2G applications.