Using Artificial Neural Networks to Predict Solar Radiation for Duhok City, Iraq

Abstract

The amount of solar radiation received at the Earth’s surface is influenced by local weather conditions. This paper investigates the effects of meteorological parameters on daily average solar radiation (DASR) in Duhok city, Iraq. Artificial Neural Networks (ANNs) based on multilayer preceptor feed-forward (MLP-FF) techniques are used to predict daily average solar radiation (DASR). The input variables used are a daily average of the relative humidity (RH), minimum temperature (Tmin), maximum temperature (Tmax), wind speed (WS), cloud layer (CL), atmospheric pressure (AP) and ultraviolet (UV) levels to estimate DASR. To identify and evaluate the effects of various input parameters on solar radiation, eight ANN-based models have been developed. To obtain the best estimation results, the number of neurons in the hidden layer has been varied. The best values of the Root Mean Square Error (RMSE), Mean Absolute Error (MAE) and correlation coefficient (R) have been calculated. For some models, the results obtained show good and better predictive accuracy than others. The present study indicates that various of the meteorological parameters can have a significant effect on the forecasting of solar radiation.