Abstract:State of charge (SOC) is an essential parameter of the battery control strategy and management system. In view of the fact that the error accumulation can not be reduced by using the integration method and the voltage method in SOC estimation of lithium battery, an algorithm has thus been proposed based on the fusion of square-root high-degree extended Kalman filter (SHEKF) and grep prediction model (GPM) to estimate SOC of lithium battery. The proposed method combines the forgetting factor recursive least square (FFRLS) method with the second-order RC equivalent circuit model to identify and modify the parameters of the lithium battery model in real time. Combined with SHEKF-GPM fusion model, an estimation had been made of the linear and nonlinear part of SOC state equation of lithium battery. The simulation results show that the error of SHEKF-GPM fusion algorithm in SOC estimation is less than 0.3%, with the covariance error being about 0%. The simulation results show that the method can reduce the error accumulation and improve the practicability, effectiveness and accuracy of battery management system in SOC estimation of lithium batteries.