The State-of-Charge (SOC) is an important evaluation index for power battery systems in electric vehicles. To eliminate the effects of drift noise in the current sensor, a new working model that takes the drift current as a state variable is proposed for high-power Li-ion batteries. In conjunction with this result, a total available capacity expression that involves the temperature, charge-discharge rate, and running mileage as variables is reconstructed by the actual operation data to improve the model accuracy for application to electric vehicles. Then, to suppress the parameter perturbations of the working model, the Unscented Particle Filter (UPF) method is applied to estimate the SOC. Experiments and numerical simulations are conducted to verify the superiority of the working model and the UPF method. The results show that the UPF method based on the working model can improve the accuracy and the robustness of the SOC estimation. (C) 2012 Elsevier Ltd. All rights reserved.