Estimation of Incomplete Meteorological and Solar Radiation Data in Turkey Using Artificial Neural Network

In this study, an Artificial Neural Network (ANN) is presented to estimate the solar radiation of specified cities (Samsun, Yozgat, Sivas, Gaziantep, Van and Adana) in Turkey. But, some of monthly values regarding these cities could be incomplete data due to unforeseen reasons. Therefore, initially these incomplete values are completed from existing data by using ANN. In order to train the ANN, the geographical and meteorological data are used in last decade (2001-2010) for the mentioned cities of Turkey. Latitude, longitude, altitude, month, year, average temperature, relative humidity, pressure, sunny day, cloudy day, and sunshine hour data are used in the input layer of ANN. The solar radiation is the output layer of ANN. The back propagation learning algorithm with two different variants, so-called Levenberg–Marguardt (LM) and scaled conjugate gradient (SCG), and tangent sigmoid transfer function are used in the network so that the best approach can be determined. After the training, absolute fraction of variance (R²) value is found as 99.84%. It is concluded that ANNs can be used to compute incomplete data and to estimate solar radiation.