Estimating the Annual Global Solar Radiation In Three Jordanian Cities by Using Air Temperature Data

Abstract

Estimating solar radiation is an imperative requirement for solar energy development in Jordan. In this paper, a quantitative approach, based on Artifiial Neural Network, was developed for estimating the annual global solar radiation of three Jordanian cities: Amman, Irbid and Aqaba. Thse cities are currently witnessing huge development and increasing demand for energy supply. Using a set of known meteorological parameters, two Artifiial Neural Network (ANN) models with diffrent architectures, called case 1 and case 2, fed with three types of learning algorithms for data training and testing, were designed to identify the optimum conditions for obtaining reliable and accurate prediction of the solar radiation. Th results showed that model case 1 performed generally better in terms of predicting the annual GSR (96%) compared to model case 2 (95%). Furthermore, the algorithms LM and SCG in general, ensured the highest effiency in training and testing the data in the designed models compared to the GDX algorithm. Threfore, model case 1, designed with one of these two algorithms, is selected as the optimal model design that is able to compute with high accuracy the annual solar radiation for the three studied cities.