Prediction of daily global horizontal solar irradiation using artificial neural networks and commonly measured meteorological parameters

Abstract

In the present study, we are interested in the estimation of the daily global solar irradiation of a horizontal surface (DGHI) based on several commonly measured meteorological variables (temperature, relative air humidity, wind speed). We test two types of Artificial Neural Networks (ANNs): Multilayer Perceptron (MLP) which is a feedforward architecture and a Nonlinear AutoRegressive neural network with eXogenous inputs (NARX) which is a recurrent architecture. The two architectures were optimized by choosing the best combination of the input parameter and by optimizing the number of hidden neurons. Results show that for all models the RRMSE is less than 20% and the coefficient of determination is greater than 97%.