Battery state of charge estimation based on multi-model fusion

The estimation of battery state of charge (SOC) is an important index of battery management system (BMS). Accurate prediction of SOC value is of great significance for the use of electric vehicles, aerospace, smart grid and even electronic products. This paper first extends the power characteristic based on the three characteristics of current, voltage and temperature. Then, based on Stacking method, the Support Sector Regression (SVR) algorithm, the Adaptive Boosting algorithm (AdaBoost) and the Random Forest algorithm (RF) are fused as the basic model, and the Linear Regression algorithm is taken as the meta-model. Finally, we use the fusion model proposed in this paper to predict battery SOC. The simulation results show that the proposed fusion model is superior to SVR, AdaBoost and RF models in the application of battery SOC prediction.