Analysis of meteorological data by artificial intelligence techniques and estimation of global solar radiation

In addition to being the main source of life in the earth, the Sun planet has an endless source of renewable energy. Turkey is situated at the sunny belt of the earth i.e. 36−42° North latitudes and 26−45° East longitudes. The global irradiation measurement used in the calculation of solar thermal or electrical energy systems is very significant for the facility location in our country. The global radiation value is equal to the sum of the beam and the diffuse rays of the radiation from the sun. Its measurement is not as easy as the measurement and the estimation of other meteorological data (such as air temperature, humidity) and long-terms are required. In the literature, various corelations have been derived by use of meteorological radiation data and the predictions for distinct geographic locations were conducted. However, it is seen that the predictions made with these corelations involve high deviations. In this thesis, meteorological data of various regions in Turkey were taken from Meteonorm software and global solar radiation forecasts were made with the proposed Artificial Neural Networks (ANN) and Support Vector Machine (SVM) by optimized the Genetic Algorithm. Generated data from the software were divided into training and testing data sets, and the proposed ANN and SVM models were tested with the data. Better results were obtained from the proposed ANN model as compared with the SVM and the existing models in the literature. Mean average percentage error was obtained to be 3.44% for testing data by the proposed ANN model which utilizes the most relevant input variables determined by the genetic algorithm.