After analyzing the strong randomness of the observed wind speed during January, April, July and October 2011 both in Hebei and Inner Mongolia, a nowcasting model of wind speed based on mean generation function (MGF), gray relational analysis and multivariate linear regression is presented, which is used to predict the future 0-4 h wind speed every 15 min. Taking the cycle of wind speed and the closed cycles between the predictors and predictand, the parameters of the nowcasting model based on MGF are determined. Based on the observed data from the wind tower in Hebei and Inner Mongolia, the applicability of the nowcasting model in wind speed is elaborated. The results show that the forecast accuracy of the nowcasting model based on MGF is improved significantly compared with the performance of the traditional autoregressive moving average (ARMA) model. Therefore, the model itself has a high value of practical application.