No abstract

This paper describes a model based on artificial neural network (ANN) for the estimation of global solar radiation (GSR) on a horizontal surface. The GSR is the primary renewable energy in the nature. The estimation of GSR is decisive; there are many researches that need such estimation since they do not have the required instruments. A simple algorithm with ANN modeling is proposed. The configuration of the back propagation neural network giving the mean square error was a three-layer ANN (5-30-1 neurons) with tangent sigmoid transfer function at the hidden layer, sigmoid transfer function at output layer, and Levenberg-Marquardt algorithm. The sensitivity analysis was developed and showed that all studied variables (time, temperature, relativity humidity, wind speed, and atmospheric pressure) have effect in the prediction of GSR. The results showed that the ANN modeling could simulate the behavior of GSR.