DYNAMIC EVALUATION OF GLOBAL SOLAR POTENTIAL IN THE REGION OF KARA (TOGO) BY ARTIFICIAL NEURAL NETWORK

In this paper, a Multilayer Perceptron model of Artificial Neural Networks with a Levenberg-Marquardt learning algorithm is designed for the prediction of solar potential in the region of Kara. For the most efficient architecture after learning, the correlation coefficient (R) and the Root Mean Square Error (RMSE) are respectively 0.84902 and 0.4000. In order to determine the model of the network to be used for the prediction, the number of neurons in the hidden layer as well as the couples of transfer functions are varied. The best results at the end of the prediction are obtained with an architecture which activation function pair is Tansig-Purelin and three (03) neurons under the hidden layer. The quality of the results obtained is satisfactory. It provides a powerful model for day-to-day prediction.