Monografías del Seminario Matemático García de Galdeano 33, 293–300 (2006) A MLP NEURAL NETWORK TO PREDICT THE WIND SPEED AND DIRECTION AT ZARAGOZA

Abstract. In this work a Multi- Layer Perceptron (MLP) Neural Network to predict the wind direction and speed at Zaragoza is introduced. The model predicts there two variables, wind speed and wind direction, in an instant t from the TEMP and SYNOP data obtained in instants t − 6, t − 12, t − 18 and t − 24 hours before. One physical-mathematical wind prediction model was showed in [2] where two Neural Networks were presented, the “time expert ” and the “spatial expert”. Here, the TEMP data classification given in [5] and [6] is added to the “time expert”. The Neural Network presented in this work, has only one hidden layer with four nodes. The outputs of model are compared with the real data obtained at Zaragoza