The energy produced from the grid connected photovoltaic system using artificial neural network prediction

The subject of this thesis is the estimation of the energy to be obtained from photovoltaic (PV) systems increasingly used to meet today's energy demand. At the beginning of the study, the power characteristics, working principle, structure of PV systems and angular relationships between PV panels and the sun as well as the radiation calculations are demonstrated with researches. Then, the basic information about artificial neural networks (ANN) and their architectures is provided for estimation of the energy to be obtained from PV systems, and the importance of normalization rules on the performance of ANN is mentioned. The aim of this study is to attempt to predict the daily amount of energy produced by a photovoltaic system of 2.3 kW in Elazig with ANN through 2013 meteorological data of Elazig. Meteorological data obtained was converted into suitable data for the artificial neural network through a series of calculations and conversion methods. Further, the data of energy amount given to the network by photovoltaic system in 2013 was obtained through inverter and used in the training phase of ANN. While data was separated into two groups as training and test data, the artificial neural network was trained by using training data and different ANN architectures. At the end of training, the test data was tried on the basis of the ANN model with the best performance and simulation data was compared with the inverter actual data. The Matlab software development environment was used in calculations and analyses during the study.