Wind energy is a highly variable energy resource. As the wind capacity of a utility grows, it becomes more difficult to effectively integrate this variable resource into the power mix. Wind power forecasting can significantly improve that integration on two levels. First, short term forecasts allow grid operators to optimize the power mix to assure reliability and stability of power transmission. If they know in advance when wind ramps are likely to occur, they can plan other resources around the wind power availability. Second, the economics of the power mix in the United States is highly contingent on day-ahead energy trading. For utilities to optimize the economic value of their power, it is critical to have a good estimate of the wind for this day-ahead period, which becomes several days ahead for weekends and holidays.

Wind power forecasting requires both Numerical Weather Prediction (NWP) simulations and statistical postprocessing capabilities. There are various ways to accomplish both. This paper will review the state-of-the-science with a focus on the post-processing technologies, which include applying model output statistics to the NWP forecasts, integrating multiple models into the forecasts, applying statistical learning or computational intelligence approaches, and accomplishing wind to power conversion. In addition, many utilities prefer probabilistic forecasts; thus some basic approaches to quantifying the uncertainty of the wind power forecasts will also be presented.