Based on the aggregative model CALMET coupled with WRF and performance parameters of the wind turbine, this paper set up the wind turbine icing forecast model by using the Makkonen icing growth model coupled with the ice surface radiation melting model. It could be used to forecast the starting and ending times of wind turbine icing. The wind turbine icing has happened 19 times from December 2012 to December 2013 in Jiugongshan Wind Farm. As the research object, three kinds of prediction schemes were used to forecast the starting and ending time of wind turbine icing, in which forecasts were made one day, two days and three days ahead, respectively. It is found that this model has certain prediction ability for wind turbine icing, and the forecasting accuracy of ice starting time was higher when closer to the ice starting time.