Predicting probabilistic wind power generation using nonparametric techniques

Wind power is becoming one of the most interesting and promising alternative for clean generation of electrical energy. The incorporation of this alternative source of energy within existing electric power system generates a series of challenges for their optimal operation. For such, reliable and accurate wind energy forecasting is required. A great deal of literature has been dedicated to this task, the majority of them departures from point forecasts to the wind speed which produces the corresponding energy point forecast using the plant wind power curve. Such methods do not take into account the uncertainty associated with wind speed. This paper proposes an alternative approach to generate wind energy forecasts, by developing a full probabilistic density forecast for the wind power for each wind speed predicted by time series methods for each lead time, using Double Seasonal Holt Winters and conditional density kernel estimation. The method was tested with real data from a Brazilian wind farm and the results obtained were very promising.