Abstract: Alleviating power fluctuation of wind farms is conducive to promoting the friendly operation of the grid, and wake control is one of the most effective measures to reduce the power fluctuation. Most existing studies are carried out based on steady wake model without considering the delay characteristics of wakes. However, wake delay can lead to the volatility of wind farm power in addition to wind uncertainty. Therefore, a quasi-steady wake model of wind farms is constructed to account for both wake interference and the delay characteristic by integrating the delay into the steady wake model. On this basis, a model predictive control considering delay (MPC-D) is proposed, and wind turbines are coordinated with the goal of command tracking and power fluctuation minimization. Finally, a medium-fidelity simulation model of a wind farm with 33 turbines is constructed on WFSim, and the impact of wake delay on the performance of wind turbines and the wind farm is analyzed. The results show that the proposed quasi-steady wake model can simulate the wake velocity loss, power delay of the turbines and power step change of the wind farm simultaneously. Moreover, the average relative error, root mean square error and sliding root mean square error of the wind farm power obtained by MPC-D are all smaller compared with those of the steady model based control. Meanwhile, MPC-D can further prevent frequent regulation of pitch angle of turbines.