Potential of adaptive neuro-fuzzy system for prediction of daily global solar radiation by day of the year Kasra Mohammadi a,⇑ , Shahaboddin Shamshirband b,⇑ , Chong Wen Tong c , Khubaib Amjad Alam d , Dalibor Petkovic ´ e a Faculty of Mechanical Engineering, University of Kashan, Kashan, Iran b Department of Computer System and Technology, Faculty of Computer Science and Information Technology, University of Malaya, 50603 Kuala Lumpur, Malaysia c Department of Mechanical Engineering, Faculty of Engineering, University of Malaya, 50603 Kuala Lumpur, Malaysia d Department of Software Engineering, Faculty of Computer Science and Information Technology, University of Malaya, 50603 Kuala Lumpur, Malaysia e University of Niš, Faculty of Mechanical Engineering, Department for Mechatronics and Control, Aleksandra Medvedeva 14, 18000 Niš, Serbia article info Article history: Received 8 October 2014 Accepted 10 January 2015 Available online 4 February 2015 Keywords: Global solar radiation estimation ANFIS technique Day of year Empirical models abstract Estimating the horizontal global solar radiation by day of the year (n day ) is particularly appealing since there is no need to any specific meteorological input data or even pre-calculation analysis. In this study, an intelligent optimization scheme based upon the adaptive neuro-fuzzy inference system (ANFIS) was applied to develop a model for estimation of daily horizontal global solar radiation using n day as the only input. The chief goal was identifying the suitability of ANFIS technique for this aim. Long-term measured data for Iranian city of Tabass was used to train and test the ANFIS model. The statistical results verified that the ANFIS model provides accurate and reliable predictions. Making comparisons with the predic- tions of six day of the year-based empirical models revealed the superiority of ANFIS model. For the ANFIS model, the mean absolute percentage error, mean absolute bias error, root mean square error and corre- lation coefficient were 3.9569%, 0.6911 MJ/m 2 , 0.8917 MJ/m 2 and 0.9908, respectively. Also, the daily bias errors between the ANFIS predictions and measured data fell in the favorable range of –3 to 3 MJ/m 2 . In a nutshell, the survey results highly encouraged the application of ANFIS to estimate daily horizontal global solar radiation using only n day . Ó 2015 Elsevier Ltd. All rights reserved. 1. Introduction Solar energy has been recognized in recent years as a proper alternative energy source owing to its unique characteristics which is free, environmental friendly and broadly accessible in most loca- tions across the globe [1,2]. In this regard, countries are performing tremendous efforts to assign a high priority to solar energy har- nessing. Nevertheless, prior to devoting any attempt for each pro- ject, accessibility to reliable and accurate solar radiation information is highly essential for experts to install and design solar energy technologies effectively [3–6]. In fact, lack of precise information on solar radiation has been a fundamental limitation in development of solar energy applications. Therefore, over the previous decades, developing suitable models and techniques to estimate the horizontal global solar radiation has been the primary objective of many researches. For this aim, a considerable number of meteorological and geographical parameters have been utilized as input elements to estimate the horizontal global solar radiation via proposed models [7–19]. Basically, despite the great history of above-mentioned empirical models as well as their relatively accu- rate global solar radiation estimates for many locations, the major deficiency of such models is their dependency to one or more cer- tain meteorological or geographical input parameters. In addition to the above-mentioned models, developing and establishing some simple models to estimate daily horizontal global solar radiation based upon day of the year as the sole input element have been the focus of some investigations in recent years [20–26]. These day of the year-based (DYB) models enjoy two significant advantages. The first merit of them is that they can be conveniently applied as there is no need to use any specific input element. Furthermore, unlike the other empirical models, applying the DYB models does not require any pre-calculation analysis such as calculating some parameters including the maximum possible sunshine duration or extraterrestrial solar radiation. In the realm of DYB models, a sine wave model was proposed by Bulut and Büyükalaca [20] to estimate the global solar radiation on http://dx.doi.org/10.1016/j.enconman.2015.01.021 0196-8904/Ó 2015 Elsevier Ltd. All rights reserved. ⇑ Correspondent authors. Tel.: +98 9121955642 (K. Mohammadi), +60 146266763 (S. Shamshirband). E-mail addresses: kasra278@yahoo.com (K. Mohammadi), shamshirband@um. edu.my (S. Shamshirband). Energy Conversion and Management 93 (2015) 406–413 Contents lists available at ScienceDirect Energy Conversion and Management journal homepage: www.elsevier.com/locate/enconman