Accurate battery state of charge (SOC) estimation is one of the most significant functions in the battery management system (BMS). Estimating the SOC of lithium battery based on the traditional unscented Kalman filter (UKF) algorithm has better estimation precision, but the application condition of the algorithm is that the statistical properties of the state process noises and measurement noises are known clearly. This paper proposed the dual adaptive unscented Kalman filter (DAUKF) algorithm which is combining the traditional Kalman filter (KF) algorithm and the adaptive unscented Kalman filter (AUKF) algorithm. Firstly, a second-order Resistor-Capacitor (RC) model of lithium battery is established; the KF algorithm is used to identify the model parameters online. Secondly, the AUKF algorithm based on Unscented Transformation (UT) is utilized to estimate the SOC. And then the implementation steps of the DAUKF algorithm are introduced in detail. The experimental results indicate that the DAUKF can estimate the SOC with error less than 1.0% under the conditions of the constant current discharge, the constant current charge and discharge and the urban dynamometer driving schedule (UDDS) cycle. Compared with the dual unscented Kalman filter (DUKF), the DAUKF has stronger estimation precision and better adaptive tracking ability.