Modeling of solar energy potential in Africa using an artificial neural network

Abstract

In this study, the feasibility of an artificial neural network (ANN) based model for the prediction of solar energy potential in Africa was investigated. Standard multilayered, feed-forward, back-propagation neural networks with different architecture were designed using NeuroSolutions®. Geographical and meteorological data of 172 locations in Africa for the period of 22 years (1983-2005) were obtained from NASA geo-satellite database. The input data (geographical and meteorological parameters) to the network includes: latitude, longitude, altitude, month, mean sunshine duration, mean temperature, and relative humidity while the solar radiation intensity was used as the output of the network. The results showed that after sufficient training sessions, the predicted and the actual values of solar energy potential had Mean Square Errors (MSE) that ranged between 0.002 - 0.004, thus suggesting a high reliability of the model for evaluation of solar radiation in locations where solar radiation data are not available in Africa. The predicted and actual values of solar energy potential were given in form of monthly maps. The solar radiation potential (actual and ANN predicted) in northern Africa (region above the equator) and the southern Africa (region below the equator) for the period of April – September ranged respectively from 5.0 - 7.5 and 3.5 - 5.5 kW h/m2/day while for the period of October – March ranged respectively from 2.5 – 5.5 and 5.5 - 7.5 kW h/m2/day. This study has shown that ANN based model can accurately predict solar radiation potential in Africa.