Global solar radiation prediction for Makurdi, Nigeria, using neural networks ensemble

Abstract

In this work, various artificial neural networks (ANNs) used for predicting the solar radiation in Makurdi city, Nigeria (7°7ʹ N long. 8°6ʹ) have been developed: Feed-forward back-propagation neural network (FFNN), radial basis function network (RBFN), and generalized regression neural network (GRNN). The main objective of this study is the use of an ANN’s ensemble for the prediction of average monthly global solar radiation of Makurdi in order to improve the prediction accuracy. The training and testing data were obtained from the Nigeria metrological station (NIMET) Makurdi. For each neural network type, average R2 = 0.998 and MSE = 0.0142, performance measures were obtained when the networks were analyzed. In order to improve prediction accuracy, an ensemble of neural networks was examined which gave an R2 = 1.0 and MSE = 0.0139. All the proposed neural networks predicted solar radiation with great accuracy; nevertheless, ensemble achieved better results implying high dependency of the model for the evaluation of solar radiation in the locations where solar radiation instruments are not available or faulty.