A Local Cascade Ensemble Learning Method for Lithium Ion Battery SOC Estimation under Multi External Factors Considering OCV Hysteresis

In order to solve the problems of hard adaptability and poor dynamic response of traditional state of charge (SoC) estimation methods, data-driven methods are adopted. In order to reduce the variance and deviation in the estimation process and make full use of the respective merits of bagging and boosting, a data-driven local cascade ensemble learning method is proposed to battery SoC estimation. Firstly, to achieve data-driven SoC estimation, the relationship between open circuit voltage (OCV) and SoC is analyzed. The effect of OCV hysteresis on SoC is discussed. Then, for the traditional SoC estimation methods difficult to consider hysteresis effect, a data-driven ensemble learning model is established to estimate SoC using relationship between OCV and SoC. Meanwhile, the form of local cascade based on random and extreme gradient boosting is adopted to minimize variance and deviation of estimation. Finally, the effectiveness and adaptability of the proposed method is verified by dataset from the built experimental bench under different external factors (temperature effect and mechanical stress). The results demonstrate that the proposed method can accurately estimate SoC under different external factors.