18. An Artificial Intelligence Approach to Forecast Wind Speeds in Algeria

Wind speed forecasts are needed for a variety of applications, like satellite launching, aviation, planning of construction, prediction of output power from wind turbines, etc. In this study, an artificial intelligence approach, namely Artificial Neural Networks (ANNs) has been applied to predict the distribution of daily wind speed at height of the 10 m of the target location using the data of neighboring locations. The study includes five regions situated at the southeast of Algeria. The data of four regions (namely Biskra, Ghardaia, Ouargla and Khenchela) have been used for training the ANN model. Thereafter, the wind speed distribution in El Oued region has been predicted. The logistic sigmoid transfer function has been used as an activation function for the hidden layer. The neurons number in the hidden layer has been selected numerically using several tests. A comparison between predicted results and actual data from NASA showed good agreements, in which the root mean square error (RMSE) is less than 1.