A novel modeling methodology for hysteresis characteristic and state-of-charge estimation of LiFePO4 batteries

Accurately estimating the State of Charge (SOC) is of utmost importance for ensuring battery safety and reliability, as well as enabling other state estimations. However, predicting the hysteresis characteristic and determining the open circuit voltage (OCV) of LiFePO4 batteries under complex charge-discharge conditions has remained a challenging task, significantly affecting SOC estimation accuracy. To overcome the limitations of existing models in describing the hysteresis characteristic of LiFePO4 batteries, this paper proposes an OCV estimation method based on the Back Propagation (BP) neural network. By integrating this approach with an equivalent circuit model, a comprehensive battery model that accurately captures the hysteresis characteristic is developed. Subsequently, the proposed battery model is combined with the Adaptive Extended Kalman Filter (AEKF) algorithm to achieve precise SOC estimation for LiFePO4 batteries. Experimental results demonstrate that the proposed method successfully reduces the hysteresis voltage estimation error to 2.5 mV, with a maximum SOC estimation error of only 1.03 %. The significance of this research lies in its ability to address the critical challenge of accurately describing the hysteresis characteristic of LiFePO4 batteries, thereby improving SOC estimation accuracy under complex operational conditions.