This paper presents the preliminary results of setting up an artificial neural network (ANN) of the feed forward type with the backpropagation training method for forecast wind speed in the region in the Isthmus of Tehuantepec, Oaxaca, Mexico. The database used covers the years from June 2008 - November 2011, and was supplied by a meteorological station located at the Isthmus University campus Tehuantepec. The experiments were done using the following variables: wind speed, pressure, temperature and date. At the same time were done seven tests combining these variables, comparing their mean square error (MSE) and coefficient correlation r, with data the predicting and experimental. In this research, is proposed a ANN of two hidden layers, for a forecast of 48 hours.