Determination of wind potential in Van and Hakkari regions

In this study, ANN (Artificial Neural Networks), Deep Learning and WASP (Wind Atlas Application Program) have been used to determine wind potentials of Van and Hakkari provinces located in eastern Anatolia region of Turkey. Data measured for 5 years period by GDM (General Directorate of Meteorology) were used for predictions of the ANN method and WASP. In the Deep Learning phase, Eastern European satellite images of the Meteosat 7 satellite were used. Using images obtained from the satellite and the data obtained from the GDM, it was attempted to estimate the target station data by the Deep Learning and ANN methods. The differences between the measured values and the estimated values of the target station are generally less than 1 m/s. It is worth to mention that, these values have excellent results for potential turbines operating under the wind speed ranging from 3 m/s to 25 m/s. A new test method, MCE (Mean Categorized Error), was presented to determine the error rate for ANN and Deep Learning Methods in the study, which determines the total error rate according to the weighted effect of each estimation of the data set. The measured data of the GDM were simulated using the WASP with the existing maps of the regions taken from the SRTM (Shuttle Radar Topography Mission) database and hence the wind potentials of the evaluated regions were determined.