short term wind speed forecasting using artificial neural network (case study: jiroft synoptic weather station)

wind speed is one of the major parameters required for an estimation of evapotranspiration and determination of crop water requirements. hence, several models and methods have been developed for a prediction of this needed climatic variable. in recent years, by development of soft computing tools, such intelligent systems as artificial neural network (ann) approach have been widely employed in agrometeorological studies. in this study, three types of four layers ann models of different number of neurons were generated and utilized for a prediction of wind speed, using hourly data of jiroft agrometeorological station, during a 6 month period, april to september, 2010. during these months winds are of higher speeds than those during the rest of the year. statistical indices of rmse, me and ef (efficiency factor) were utilized for comparisons and as well for models' evaluations. the results revealed that an ann model with 20 neurons in each layer is of the most suitable performance in prediction of wind speed with the respective corresponding values of these indices as 1.1827, 0.6947 and 0.9246.