This paper proposes a dual extended Kalman filtering ( DEKF) algorithm for estimating the State-of-Charge ( SOC) of lithium-ion batteries on line. First of all, the state-space representation of the battery model is established based on Thevenin battery model and Kalman filtering algorithm. The least squares method and the DEKF algorithm are used to identify the battery model parameters，which improves the model accuracy and facilitates the battery model to well reflect the actual internal state of the battery, Moreover， the principle of using DEKF algorithm to estimate the inner SOC of the battery on line is introduced，and corresponding battery test experiments are designed. Experiment results demonstrate that under various operating conditions, the algorithm has relatively high accuracy and good environment adaptability when applied to evaluate SOC on line; and the maximum error is less than 4.5%. The DEKF algorithm is proved to have good convergence and robustness，and can efficiently solve the problems of inaccurate initial- value estimation and error accumulation.