Global solar radiation estimation using artificial neural network by the addition of nearby meteorological stations' solar radiation data and exergy of solar radiation: a case study

Abstract

The artificial neural networks (ANNs) can be used to accurately predict the global solar radiation (GSR). There are many geographical, meteorological and terrestrial parameters affecting GSR. In this study, the most relevant of six input parameters are selected to predict the GSR of Goksun Station in Turkey using Waikato environment for knowledge analysis (Weka) Software. The effect of using nearby meteorological stations' GSR data as input on GSR prediction is investigated. Different ANN models are developed to demonstrate the difference between the exclusion and inclusion of these parameters on the model. The results show that the exclusion of less influential parameters and the inclusion of three nearby stations' GSR data has improved performance criteria. Petela, Spanner and Jeter's approaches are used for exergy analysis of measured and estimated GSR values. The mean exergy-to-energy ratio for both Petela and Spanner's approaches is 0.934, while Jeter's approach showed 0.950.