A dynamic weighting ensemble approach for wind energy production prediction

In this work, we propose a method to predict wind energy production. The method is based on an ensemble of Artificial Neural Networks (ANNs), which receive in input weather forecast variables and predict the wind plant energy production. We investigate different strategies for aggregating the outcomes of the individual models of the ensemble and compare them with a real dataset. A dynamic weighting ensemble which combines the individual models outcomes proportionally to their local performances in the neighborhood of the test pattern under analysis is found to provide the most accurate predictions.