Considering the required fixed-point and hourly wind speed forecast for the wind power prediction in wind farm,numerical prediction of wind velocity from the MM5 lattice point output released by the Central Meteorological station was interpolated into of an anemometer tower at a certain height in a wind farm along the coast of Fujian in this paper.Through error analysis of forecast results,we found that there were some systematic errors due to special underlying surface at the junction of land and sea.According to the lag correlation of the error and the real-time meteorological data from the anemometer tower,weinvestigated dynamic correction methods for numerical prediction of wind velocity from the MM5 using earlier error observations and anemometer tower turbulent indicators,and established revised equations.The average absolute error of the revised wind velocity forecast was reduced by 31%—54% and the accuracy of prediction was improved effectively.