Abstract: A sliding mode observer (SMO) algorithm based on joint Extend Kalman Filter (EKF) was proposed for state of charge (SOC) estimation of electric vehicle batteries. The second-order Thevenin equivalent circuit model was used to describe the characters of the battery, and its parameters at different temperatures were identified. Based on the results, the influence of temperature on the parameters and accuracy of the battery model were analyzed. Aiming at the high dependence of EKF on model accuracy and the serious chattering of estimation results caused by the sensitivity of SMO to noise, a new SOC estimation algorithm of combining sliding mode observer with EKF was proposed, in which a chattering reduction function is added to the state correction equation of EKF and the relevant parameters of the function are obtained according to the stability constraints of SMO. The proposed algorithm can synthesize the advantages of EKF and SMO at the same time, and has strong robustness to the modeling error while filtering noise. Finally, the corresponding simulation conditions were designed and a series of experiments have been carried out. The experimental results show that the proposed algorithm has higher estimation accuracy than both EKF and SMO in complex vehicle operating conditions.