The aim of this paper is to present an advance of the study carried out to develop a Physical-Mathematical wind prediction model in some zones of the Ebro Valley. This study is divided into several phases. It begins with the capture, filtering and descriptive study of the necessary series of data. Next, some of the variables of interest are fitted to known probability distributions, such as Weibull or Exponential. In the last part, neural networks techniques are used on the series of data. Mainly, two related networks are used. The first one studies "time" prediction, estimating the velocity and the direction of the wind in a place after a lapse of time. The other network studies "spatial" prediction, estimating the same variables in different places at the same instant of time.