In view of estimating the state of charge (SOC) of lithium batteries in electric vehicles, SOC estimation algorithm of a ternary lithium battery based on improved recursive least squares method is proposed. Firstly, a first-order RC equivalent circuit model of a ternary lithium battery is established. Then, based on the recursive least squares method, the dynamic constraints and backtracking of the parameters to be evaluated are considered, the model parameters are identified online, and extended Kalman filter (EKF) is combined to estimate SOC. Finally, a dynamic operating condition simulation test is performed on the ternary lithium battery. The results show that the improved recursive least squares method can effectively suppress the abnormal jitter of the parameters in complex operating conditions, and reduce the terminal voltage prediction error. Combining the method with EKF to estimate SOC can achieve higher accuracy.