Comparative Study of Joint Estimation of State of Charge (SOC) and State of Health (SOH) of Lithium-ion Batteries Based on Different Tree Models

The realization of accurate State of Health (SOH) and State of Charge (SOC) estimation is a prerequisite to ensure the safe use of energy storage batteries, which helps to further improve the energy utilization efficiency effectively. Data-driven methods are efficient, accurate, and do not rely on accurate battery models, which is a hot direction in battery state estimation research. However, the relationships between variables in the lithium-ion battery dataset are mostly nonlinear, which largely affects the prediction of the model. In addition, the model also has a series of defects, such as large computation, strong data dependence, and long consumption time. In this paper, a joint online estimation method of battery SOC-SOH based on tree modeling algorithm is proposed to solve the above problems. Based on NASA battery sample data, this study explores the changing law between SOC and discharge voltage and temperature under different State of Health (SOH). Subsequently, a combination of RFR, GBDT and XGBoost tree modeling algorithms are used for battery SOC-SOH estimation based on the above variation rules. The experimental results show that the R2 scores of the XGBoost algorithm in predicting both SOC and SOH are more than 0.995, indicating its good adaptability and feasibility.