Because the wind generation has a quick starting speed and needs small cranking power, it can be employed to accelerate the restoration speed of blackout system. Because of the uncertainty of wind power output, the output power of wind farm must be scheduled to reduce the fluctuation and ensure the security of restored system, as well as enhance the efficiency of wind power utilization during system restoration. In this paper, an information-gap decision theory (IGDT) based optimization model is proposed to dispatch the power of wind farms during system restoration. Firstly, the deterministic model is established without considering the wind power uncertainty. Then the deterministic model is transferred to the robust model by IGDT considering the uncertainty of wind power, which is solved by artificial bee colony method. Lastly, the method is verified in the New England system and part of Jiangsu power grid.