Under the background of large-scale electric vehicles(EV)integrated into the power grid, in order to coordinate and control the charging and discharging power of EV group,so as to make EV a useful energy storage element which can provide frequency regulation for the power grid, the self-organizing classification method of EV group from the aspect of state-of-charge(SOC)is studied in this paper, and a self-organizing map (SOM)neural network is designed to divide the EV group into several classes in the mathematical space according to the SOC.Based on the clustering results and the SOC level of each EV individuals in different classes, a control method of variable parameter droop frequency regulation for EV charging and discharging control terminal is presented.The simulation results show that by applying the proposed control method, the large-scale EVs integrated into the power grid with different SOC levels will share the peak load and surplus power of the power grid in proportion to their respective capacity, and meet the demand of power gird frequency regulation.