With the increasing of wind power permeability, largescale wind power participation in frequency modulation becomes the most important part of wind power development, and power allocation method needs to be more intelligent and simple. Therefore, the objective function optimization method is used to optimize the power allocation of wind farm layer. Firstly, the frequency modulation power function of the whole wind farm is modeled so that the realtime power change value of the whole wind farm tracks the power to be distributed in the whole wind farm, and the minimum deviation between the above two is taken as the objective function. Secondly, the upper and lower limits of wind speed and active power output of each wind turbine in a certain period are obtained by ultrashortterm prediction method, and the variation limit of the pitch angle is taken as the restriction condition. Finally, the optimized power deviation value of each wind turbine is obtained and superimposed on the basic value of each wind turbine output for frequency modulation. By comparing with other power allocation methods, the superiority of the objective function optimization method is verified, which can improve the speed and reliability of wind farm frequency modulation.