Abstract: The large-scale integration of wind power leads to a serious reduction of system inertia, which deteriorates the frequency stability of the system. For the above problems, a short-term frequency support scheme for doubly-fed wind turbines participating in grid frequency regulation is proposed based on nonlinear prediction theory and extended state observer (ESO) theory. First, the target state equation related to system frequency deviation and wind turbine speed is established, and the corresponding prediction order of different control objectives is calculated. Then, according to the nonlinear prediction theory, the prediction matrix is calculated to predict the state tracking error at the future time, and the nonlinear control law is obtained. Finally, the ESO is introduced to observe the complex Lie derivative operation in the nonlinear control law, which reduces the computational burden required for the control law. The wind power system is built by MATLAB/SIMULINK for simulation verification. The results show that the proposed nonlinear scheme can effectively improve the frequency response of the power grid, and realize the self-recovery of the wind turbine speed without designing the speed recovery link separately.