Eur. Phys. J. Plus (2022) 137: 289
https://doi.org/10.1140/epjp/s13360-022-02398-z

Regular Article

Estimating daily global solar radiation in hot semi-arid climate using an efficient hybrid intelligent system

¹ Faculty of Engineering, Shohadaye Hoveizeh Campus of Technology, Shahid Chamran University of Ahvaz, Dashte Azadegan, Ahvaz, Iran
² Department of Civil Engineering, Behbahan Khatam Alanbia University of Technology, Behbahan, Iran
³ Water Science and Engineering Dept, Ferdowsi University of Mashhad, Mashhad, Iran
⁴ Khuzestan Water and Power Authority, Ahvaz, Iran
⁵ Faculty of Agriculture, University of Zanjan, Zanjan, Iran
⁶ School of Surveying and Geospatial Engineering, College of Engineering, University of Tehran, Tehran, Iran
⁷ College of Computer Science and Artificial Intelligence, Wenzhou University, 325035, Wenzhou, Zhejiang, China

^a m.jamei@shhut.ac.ir

Received: 13 April 2021
Accepted: 17 January 2022
Published online: 1 March 2022

Abstract

Solar energy is one of the most important renewable energy sources. Assessing the area's solar potential needs analyzed information about the dataset of the measured global solar radiation (GSR). Researchers recently detected the high potential of state-of-the-art artificial intelligence (AI) methods in successfully estimating the GSR. In this study, a novel hybrid AI-based tool consisting of a least square support vector machine (LSSVM) integrated with improved simulated annealing (ISA) is proposed to predict the GSR over the Ahvaz synoptic station located in the South-West of Iran. The potential of the proposed hybrid paradigm so-called LSSVM-ISA was evaluated by using multivariate adaptive regression spline (MARS), generalization regression neural network (GRNN), and multivariate linear regression with interactions (MLRI). For precise assessment of efficiency of the AI models, various statistical metrics and validation methods were used to assess the precision of the developed models. A comparison of the obtained results indicated that the LSSVM-ISA method performed better than the MARS, GRNN, and MLRI models. The achieved RMSE values of the MARS, GRNN, and MLRI models were decreased by 9%, 16%, and 30% using the LSSVM-ISA model. Finally, the results demonstrated that the LSSVM-ISA model could be successfully employed for accurately estimating GSR.