ISESCO JOURNAL of Science and Technology
The Official Journal of ISESCO Centre for Promotion of Scientific Research


In this work, an Artificial Neural Network (ANN)- based model for solar energy prediction using climatic parameters in a tropical region (Garoua; 9.3 N, 13.4 E, altitude: 242 m) was developed. Standard multilayered, feed-forward back-propagation neural networks with different architectures were designed and programmed. Meteorological data for a nineyear period (1995-2003) from NASA’s geo-satellite database and the National Meteorological Service were used for training and testing the network. Four meteorological parameters (sunshine duration, temperature, relative humidity and pressure) and two temporal parameters (month and day), were used as inputs to the network, while the solar radiation intensity was used as the output of the network. The results of the best model show that the correlation coefficient between the ANN predictions and the actual global solar radiation (GSR) intensities is close to 98%, with a low value of mean bias error (MBE= 0,072) and root of mean square error (RMSE=0,2), thus suggesting a high reliability of the model in evaluating solar radiation. In order to test the ANN model, statistical analysis is performed using the results obtained from other models, including a linear Angström-Prescott variation, a quadratic equation, a logarithmic variation, an exponential function, and the Heliost_2 method. The model seems promising for evaluating the solar resource potential in places where there are no monitoring stations, as in poor countries.

Keywords: Artificial Neural Network, Modelling, Solar Radiation, Tropical Region.