Prediction of Monthly Average Daily Global Solar Radiation and Average Temperature of Next Month Using ANN Based on Previous Months’ Data of Some Cities in Turkey

Turkey has a good solar energy potential due to its geographical location of 36°- 42° N latitudes. In this study, an Artificial Neural Network (ANN) is developed to predict the monthly average daily global solar radiation and average temperature. The methodology is based on ANN techniques and has been applied to the meteorological data of the selected six cities, which are Mersin, Adana, Antakya, Malatya, Kayseri and Sanliurfa in Turkey. In order to train and test the ANN model, meteorological data are used for recent years (2001-2010) from six stations spreading over Turkey. Five-fold cross-validation method is utilized during training and test stages. The meteorological and geographical data are month, year, relative humidity, pressure, sunny day, cloudy day, solar radiation, average temperature and sunshine hours. Previous month’s all data are assigned as the input layer of ANN while the monthly average daily global solar radiation and average temperature values of next month are the output layer of ANN. The results obtained by ANN and the measured data are compared to each other. The average accuracy ratio of predicted values is found as %99.35 and %99.57 for average temperature and monthly average global solar radiation, respectively. Results have shown that the predicted and measured values are very close to each other. INTRODUCTİON Conventional energy sources such as coal, oil, and nuclear have negative effects global warming and environmental pollution. Therefore, researchers on renewable energy sources known as the futuristic energy resources have been studied as hot topic in recent decade. In general, solar and wind energy are considered to be good alternative energy sources for tackling with these problems due to their safety and positive contribution to the global environmental (Jacovides et al., 2002; Jacovides et al., 2006). Solar radiation data are used for variety purposes. For example, beam and diffuse solar radiation on horizontal surfaces are widely used for simulations, modeling and sizing of solar processes. The solar radiation data, which is measured as hourly and daily, has significant importance according to application area such as, pv system designs, meteorology, solar maps, and engineering applications. Total or global radiation incident on surface consists of three components which are beam radiation, diffuse radiation, and reflected radiation from the ground. Some researchers have used sky clearness to measure surface global solar irradiance (Iqbal, 1983; Meek, 1997). One of the most accurate methods to evaluate the surface solar radiation is measured by pyranometers. A pyranometer is an instrument producing a voltage as function of the incident radiation. Automatic acquisition datalogger is used for pyranometer measurements of solar irradiance (Gu L, 2001; Segal, 1992; Souza JL., 2005). Some of researchers have used Artificial Neural Networks (ANN) to predict total solar radiation (Sozen et al., 2004; Tymvios, 2005; Elminir et al., 2005; Mellit et al., 2005; Mihalakakou et al., 2000; Mohandes et al., 1998; Reddy and Ranjan, 2003; Alawi and Hinai, 1998). In addition to modeling solar radiation, ANN have been widely used in several applications such as pattern recognition and classification, function approximation and prediction, identification and control (Knutti et al., 2003; Slavisa et al., 2001). And also, many authors have reviewed and investigated parametric models to predict solar radiation (Gueymard, 1989; Gueymard, 1993; Gueymard, 1998; Gueymard, 2001; Maxwell, 1998; Muneer et al.,1998; Muneer et al.,2000). In this study, an ANN model is developed to predict the monthly average daily global solar radiation and average temperature values of next month by using previous month’s data. The predicted values are compared with the measured pyranometer and thermometer data.