The invention discloses a short-term power prediction method applied to a newly-built wind power plant, which comprises the following steps of performing downscaling operation on the prediction outputof a medium-scale mode to obtain a short-term predicted wind speed of the hub height at the position of an early-stage anemometer tower of the wind power plant, and establishing a new wind power plant short-term power prediction model according to the wind sped-power characteristic curve analytic function and the comprehensive plant power consumption ratio of the wind power plant to realize short-term prediction of the overall output of the wind power plant. According to the prediction method, the uncertainty caused by insufficient mesoscale resolution and the defect that a high-precision statistical model cannot be established due to lack of a large amount of historical operation data in the new wind power plant are effectively reduced, and the short-term power prediction precision of the new wind power plant is remarkably improved.