Comparison of two intelligent models to estimate the instantaneous global solar radiation in semi-arid climate conditions: Application in Iran

Abstract

Solar radiation incident on the earth's surface is a fundamental input for many aspects of climatology, hydrology, biology, and architecture. In addition, it is an important parameter in solar energy applications. Due to the high cost of the measuring instruments of solar radiation, many researchers have suggested different empirical methods to estimate this essential parameter. In this study, with the help of fuzzy systems and neural networks, two models have been designed to estimate the instantaneous global solar radiation in Rafsanjan city which has a typical climatic conditions of semi-arid region of middle eastern countries. In fuzzy and neural network model, the inputs are the number of the given day in the year, time, ambient temperature and cloudiness, The comparison between the results of the models and the measurements, shows that the estimated global radiation is similar to the measurement; for fuzzy model, statistical indicators RMSE, MBE and t-test are 103.4367 (w/m 2), 4.1169 (w/m 2) and 9.1318, respectively and for ANN, they are 85.46 (w/m 2), 3.08 (w/m 2) and 5.41, respectively. As the results indicate, both models are able to estimate the amount of radiation well, while the neural network has a higher accuracy. The output of the modes for six other cities of Iran, with similar climate conditions, also proves the ability of the proposed models.