Predicting global solar radiation for North Algeria

Abstract

Having a complete and effective radiometric database is very crucial in the renewable energy field for the design of solar photovoltaic and thermal system. Despite the existence of the chain of radiometric measurement at the University of Blida, data acquisition of various components of radiation is still having problems, such as gaps in basic radiometric data due to heavy power-cuts especially in summers. Thus, a good design is only possible if the measurements are available continuously in space and time. Solar energy estimation procedures using artificial neural networks methods may overcome the issue. In this study, an artificial neural network (ANN) was used for the estimation of daily global solar radiation (DGSR) on horizontal surface using data measured from the meteorological station located inside the University. Six input parameters were used to train the network. These parameters were elevation, longitude, latitude, air temperature, relative humidity, and wind speed. The optimized network obtained with lowest error during the training was one with 6 neurons in the input layer; 6 neurons in the hidden which was obtained by trial and error, and one neuron in the output layer. The results show that the ANN can be accurately trained and that the chosen architecture can estimate the DGSR with acceptable accuracy: mean absolute error (MAE) less than 20% for both training and validation step.