The state of health estimation of lithium-ion batteries based on data-driven and model fusion method

Abstract

State of health (SOH) estimation of lithium-ion batteries is widely concerned. Currently, electric vehicles are rarely complete discharging in practical application, which remains lots of electricity and reduces constant current charging time. Therefore, this phenomenon hinders the applications of many traditional methods that require a complete constant current charging process. In this paper, we put forward a data-driven and model fusion method for SOH estimation based on constant voltage charging process (CVCP). Firstly, an improved equivalent circuit model (IECM) is established based on the current-time data of the CVCP. Secondly, Pearson correlation coefficient describes the strong mapping relationship between model parameters and SOH, so the model parameters are used as health indicators. Then, SOH prediction model is established by back propagation neural network whose model parameters are optimized by improved particle swarm optimization algorithm. Thirdly, considering time-consuming problem, a new scheme based on the incomplete CVCP that combine time constants prediction model and SOH prediction model is adopted. Finally, comparative results show that proposed IECM has the higher current estimation accuracy than traditional equivalent circuit models for different batteries. The SOH maximum errors of proposed method in different temperatures and data lengths are both within 2%.