Probabilistic short-term wind power forecasting based on kernel density estimators

Short-term wind power forecasting tools have been developed for some time. The majority of such tools usually provide single-valued (spot) predictions. Such predictions are however often not adequate when the aim is decision-making under uncertainty. In that case there is a clear requirement by end-users to have additional information on the uncertainty of the predictions for performing efficiently functions such as reserves estimation, unit commitment, trading in electricity markets, a.o. In this paper, we propose a method for producing the complete predictive probability density function (PDF) for each time step of the prediction horizon based on the kernel density estimation technique. The performance of the proposed approach is demonstrated using real data from several wind farms. Comparisons to state-of-the-art methods from both outside and inside the wind power forecasting community are presented illustrating the performances of the proposed method.