With the rapid development of wind power in recent years, to achieve more accurate predictions of wind speed on wind farm has gradually become an hot issue in research. To improve the wind speed forecasting accuracy for wind farm, a new self-adaptive and dynamic forecasting method is presented, which is based on the wind speed time series and traditional cubic exponential smoothing (ES) method. According to the principle of minimum error sum of squares, the smoothing factor is chosen with blanket search method and changed in time. Then one-step or multistep wind speed forecasting can be done with the new method. By comparing with the traditional cubic exponential smoothing method and gray model forecasting method, it can be proved that the self-adaptive and dynamic forecasting method in wind speed forecasting is accurate and efficient.