Abstract: The state-of-charge(SOC) is an important parameter for the electrical vehicle. A new Ni-MH battery nonlinear dynamic model in discrete-time state-space form was introduced. In the new model, SOC and state-of-polarization as states of the state vector were introduced to represent the dynamic behavior of the battery more accurately. Taking into account the charge accumulating effect that influence the terminal voltage, the model was modified by adding an accumulating item into the state equation. The SOC estimation method based on extended Kalman filter was studied. The battery voltage, current, internal resistance and temperature were used in the algorithm. The calculation equations were presented in detail. Charge and discharge experiments with constant current, pulse current and variable currents were implemented. The simulation with these experiment data shows that the method is benefit to improve the accuracy of SOC estimation. The method can be used in a battery management system and applied in hybrid electrical vehicles.