Abstract: Evaluation of the state of charge (SOC) is a key technology for electric vehicle battery management. This work develops a method to establish the relationship among Coulomb efficiency, SOC and charge/discharge current (I). The curve of SOC to I (SOC-I) is provided that could supply a reasonable Coulomb efficiency during prediction. Moreover, the algorithm of adaptive unscented Kalman filtering (AUKF) is used for battery SOC evaluation. A new SOC-I-AUKF algorithm combined the AUKF algorithm with SOC-I curve is developed. During the process of SOC prediction, the new algorithm could adjust the Coulomb efficiency, process noise covariance and measurement noise covariance to reach the optimal evaluation. Experiment results indicate that the SOC-I-AUKF algorithm has better performance than UKF algorithm in prediction of absolute error, relative error and average error.