With the rapid development of electric vehicles, the accurate estimation for the state of charge (SOC) of power battery is crucial for vehicle energy management, range prediction and prevention of overcharge/overdischarge. In order to improve the estimation accuracy of lithium battery SOC, an equivalent circuit model was constructed based on the fractional order theory, and the model parameters were identified by using the adaptive forgetting factor recursive least square (AFFRLS) algorithm. In order to solve the problem of low data utilization and insufficient noise immunity caused by the single unscented Kalman filter (UKF) under dynamic conditions, a fractional order multiple-information unscented Kalman filter (FOMIUKF) algorithm was proposed by combining with the requirements of the long memory characteristics of fractional order model. The simulation model was built by MATLAB and compared with the extended Kalman filter (EKF) algorithm and UKF algorithm. The results showed that the average error of SOC estimation based on FOMIUKF algorithm was 0.78%, which was 0.42% higher than that of EKF algorithm and 0.25% higher than that of UKF algorithm.