Forecasting wind power is recognized as a tool in mitigating the operational challenges imposed on power systems by large‐scale integration of intermittent wind‐powered generators. Wind energy is directly dependent upon wind speed, which is a complex signal to model and forecast. In this paper, a new Hybrid Iterative Forecast Method (HIFM) for wind speed forecasting is presented which takes into account the interactions of temperature and wind speed. To select the most relevant and the less redundant input variables from the available data, a two‐stage feature selection technique is also introduced. The forecast accuracy of the proposed wind power prediction strategy is evaluated by means of real data of wind power farms of Iran and Spain's power systems. Copyright © 2010 John Wiley & Sons, Ltd.