Proceedings of the Global Conference on Global Warming 2012 8-12 July, 2012, Istanbul, Turkey - 1 - COMPARISON OF CLASSICAL REGRESSION AND ARTIFICIAL NEURAL NETWORK MODELS FOR PREDICTION OF GLOBAL SOLAR RADIATION IN DUBAI, UAE Hassan A.N. Hejase , Ali H. Assi, Maitha H. Al -Shamisi Department of Electrical Engineering at the United Arab Emirates University, PO Box 17555, Al-Ain, UAE Abstract —This study is a continuation of the ongoing effort by the authors to establish a weather model for the UAE. Previous work considered weather data for the cities of Abu Dhabi and Al-Ain in the United Arab Emirates using several modeling approaches including classical regression, Artificial Neural Network (ANN), and time-series regression models. The classical regression models used included the linear Angstrom- Prescott model and its derivations, namely, the second and third order correlations, in addition to the single term exponential model, logarithmic model, linear logarithmic model and power model. This work considers the use of empirical classical regression for predicting the monthly average daily global solar radiation in the city of Dubai, UAE. Available weather data included the mean air temperature ( 0 C), mean wind speed (knots), daily sunshine hours, and percent relative humidity in addition to the daily global solar radiation (kWh/m 2 ). The data was divided into two groups: one data group from 2002-2008 for the prediction model and the second group (2009-2010) for testing the model. Results of monthly mean GSR comparison for test period of 2009-2010 for all empirical models yielded low statistical error parameters and coefficients ofdetermination better than 96 %. Comparison with ANN multilayer preceptor (MLP) and radial basis function (RBF) models shows that the optimal ANN MLP model is the best with a coefficient of determination R 2 = 98 %, root-mean-square-error RMSE=0.22, mean absolute bias error MABE= 0.20767, mean bias error MBE= 0.04826, and mean absolute percentage error MAPE= 3.85 %. However, the empirical regression models make use only of one weather variable, namely, sunshine hours, which is advantageous as the ANN model selected uses three variables (maximum temperature, sunshine hours, relative humidity). Keywords— Global solar radiation, regression, ANN 1. INTRODUCTION Many researchers have modeled weather data using classical regression and Artificial Neural Networks (ANN) techniques. Numerous authors (Hassan A.N. Hejase and Ali H. Assi 2011; Abdalla & Feregh, 1988; Assi & Jama, 2010; Akinoglu & Ecevit, 1990; Al Mahdi et al., 1992; Ampratwum & Drovlo, 1999; Elagib & Mansell, 2000; Falayi et al., 2008; Fortin et al., 2008; Khalil & Alnajjar, 1995; Menges et al., 2006; Newland, 1988; Podesta’ et al., 2004; Sahin, 2007; Samuel, 1991; Ulgen & Hepbasli, 2002) to count few developed empirical regression models to predict the monthly average daily global solar radiation (GSR) in their region using various parameters. The mean daily sunshine duration was the most commonly used and available parameter. The most popular model was the linear model by Angström-Prescott (Podesta’ et al., 2004; Assi & Jama, 2010) which establishes a linear relationship between GSR and sunshine duration with knowledge of extra-terrestrial solar radiation and the theoretical maximum daily solar hours. Many studies with empirical regression models were done for diverse regions around the world. Menges et al. (2006) reviewed 50 GSR empirical models available in literature for computing the monthly average daily GSR on a horizontal surface. They tested the models on data recorded in Konya, Turkey for comparison of model accuracy. The number of weather parameters varied between models. The diverse regression models used include linear, logarithmic, quadratic, third order polynomial, logarithmic-linear and exponential and power models relating the normalized GSR to normalized sunshine hours. Other models included in Menges work used direct regression models involving various weather parameters such as precipitation, cloud cover, etc., in addition to geographical data (altitude, latitude). Şahin (2007) presented a novel method for estimating the solar irradiation and sunshine duration by incorporating the atmospheric effects due to extra-terrestrial solar irradiation and length of day. The author compares his model with Angström’s equation with favourable advantages as his method does not use Least Square Method in addition to having no procedural restrictions or assumptions. Ulgen and Hepbasli (2002) developed two empirical correlations to estimate the monthly average daily GSR on a horizontal surface for Izmir, Turkey. Their models resemble Angström type