The change rate of terminal voltage of a LiFePO4 battery is too small to be used to estimate state of charge (SOC) during discharging when the battery is within 10-80% SOC. A coulombic counting SOC estimation (Ah) method has the disadvantages of accumulated error and start point error. Therefore, in this paper the Extended Kalman filter (EKF) is used. The estimation accuracy of EKF deeply depends on the model's accuracy, hence a series of experiments must be performed to build an accurate model. Finally, comparisons are also made between the EKF and Ah methods. The performance of this model with the EKF method is higher than that with the Ah method.