Machine learning methods to predict wind intensity

Abstract

A decision making problem often becomes a problem of selection. In this kind of problems (decision making or fo-recasting problems) the selection of an effective set of input variables, which is usually a complex and sometimes an unmanageable process, is the main problem in real situations. The correct selection of the most important data in the assessment of a problem allows not only faster decision but the reduction of the prediction error. In this paper we use a hybrid model of a Genetic Algorithm as a heuristic tool, to select appropriate combinations of different variables that have more effect on forecasting decision making parameters, and Artificial Neural Network as a fitness function of genetic algorithm. The model was then applied to predict the intensity of the wind in the short and medium term in the central-south region of Portugal. The results proved to be excellent regardless of the forecast horizon.