Comparison of Empirical Models and an Adaptive Neuro Fuzzy Inference System for Estimating Hourly Total Solar Radiation on Horizontal Surface at Alexandria City, Egypt

Abstract

Solar radiation data in a particular location is an important factor for agricultural applications and others. To estimate solar radiation, empirical models have been developed using different meteorological parameters. Recently, prediction models based on artificial intelligence techniques such fuzzy logic are available. The aim of this work was to develop an adaptive neuro fuzzy inference system (ANFIS) for estimating hourly total solar radiation on horizontal surface at Alexandria city, Egypt and to compare its efficiency with two empirical models namely clear sky hourly global solar radiation  and global solar flux on a horizontal surface. Local time, Julian day, air temperature, relative humidity and relative sunshine duration data for the period 2005-2007 were used as inputs to ANFIS model. Delta-T automatic weather station which was located on the roof-top of Agricultural and Bio-Systems Engineering Department, Faculty of Agricultural, Alexandria, Egypt was employed to collect the required data. In testing phase, good results with all prediction methods were obtained, with root mean square error values of 165.42, 168.37 and 82.287 W/m2 for clear sky hourly global solar radiation model, global solar flux on a horizontal surface model and ANFIS model, respectively. Meanwhile, coefficients of determination (R2) were 0.6428, 0.6355 and 0.8949, respectively for clear sky hourly global solar radiation model, global solar flux on a horizontal surface model and ANFIS model when utilized testing data set for the validation process. Even though all the investigated models can be used to predict the hourly total solar radiation on horizontal surface, ANFIS model produced better estimates.