Integrating experimental and theoretical approaches for enhanced machine learning modeling of solar radiation

This study presents a novel hybrid framework for estimating solar radiation components—critical for optimizing solar energy systems—by integrating theoretical solar parameters with meteorological data, an approach seldom explored in literature. A one-year dataset, recorded at 15-minute intervals, wa...

Full description

Saved in:
Bibliographic Details
Main Authors: Nader Ghareeb, Abeer Alanazi, Ahmad Sedaghat, Mohamad Hussein Farhat, Arash Mehdizadeh, Hayder Salem, Mohammad Nazififard, Ali Mostafaeipour
Format: Article
Language:English
Published: Elsevier 2025-10-01
Series:Engineering Science and Technology, an International Journal
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2215098625002113
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849338308821254144
author Nader Ghareeb
Abeer Alanazi
Ahmad Sedaghat
Mohamad Hussein Farhat
Arash Mehdizadeh
Hayder Salem
Mohammad Nazififard
Ali Mostafaeipour
author_facet Nader Ghareeb
Abeer Alanazi
Ahmad Sedaghat
Mohamad Hussein Farhat
Arash Mehdizadeh
Hayder Salem
Mohammad Nazififard
Ali Mostafaeipour
author_sort Nader Ghareeb
collection DOAJ
description This study presents a novel hybrid framework for estimating solar radiation components—critical for optimizing solar energy systems—by integrating theoretical solar parameters with meteorological data, an approach seldom explored in literature. A one-year dataset, recorded at 15-minute intervals, was collected from a weather station at The Australian University, Kuwait. The theoretical global solar radiation values were validated against 8,784 hourly averaged experimental data points, incorporating key solar parameters such as local time, solar declination, the Equation of Time (EoT), hour angle, and solar zenith angle. These theoretical inputs, combined with meteorological parameters (temperature, humidity, wind speed, and wind direction), were used to train multiple machine learning (ML) models. To improve model accuracy, a four-scenario data treatment strategy was applied, including Gaussian fitting and zero-removal techniques. In total, 28 ML models were evaluated, encompassing linear regression, support vector machines, Gaussian process regression, and neural networks. The Exponential GPR model emerged as the most accurate, achieving R2 values > 0.99 and RMSE as low as 0.13 W/m2. These results represent a significant performance improvement over existing models in the literature, many of which report RMSE values in the range of 20–60 W/m2 and R2 values between 0.85 and 0.98. The findings demonstrate substantial improvements over existing methods and offer a scalable solution for a more precise solar forecasting in various environments.
format Article
id doaj-art-b7fac37cde124178ae4d539682568da2
institution Kabale University
issn 2215-0986
language English
publishDate 2025-10-01
publisher Elsevier
record_format Article
series Engineering Science and Technology, an International Journal
spelling doaj-art-b7fac37cde124178ae4d539682568da22025-08-20T03:44:27ZengElsevierEngineering Science and Technology, an International Journal2215-09862025-10-017010215610.1016/j.jestch.2025.102156Integrating experimental and theoretical approaches for enhanced machine learning modeling of solar radiationNader Ghareeb0Abeer Alanazi1Ahmad Sedaghat2Mohamad Hussein Farhat3Arash Mehdizadeh4Hayder Salem5Mohammad Nazififard6Ali Mostafaeipour7Department of Mechanical Engineering, College of Engineering, Australian University, Safat 13015 West Mishref, KuwaitDepartment of Mechanical Engineering, College of Engineering, Australian University, Safat 13015 West Mishref, KuwaitDepartment of Mechanical Engineering, College of Engineering, Australian University, Safat 13015 West Mishref, KuwaitDepartment of Electrical and Electronics Engineering, College of Engineering, Australian University, Safat 13015 West Mishref, KuwaitDepartment of Electrical and Electronics Engineering, College of Engineering, Australian University, Safat 13015 West Mishref, KuwaitDepartment of Mechanical Engineering, College of Engineering, Australian University, Safat 13015 West Mishref, KuwaitUniversité Côte d’Azur, Polytech’Lab, France; Corresponding author.Civil and Environmental Engineering Department, California State University, Fullerton, CA, USAThis study presents a novel hybrid framework for estimating solar radiation components—critical for optimizing solar energy systems—by integrating theoretical solar parameters with meteorological data, an approach seldom explored in literature. A one-year dataset, recorded at 15-minute intervals, was collected from a weather station at The Australian University, Kuwait. The theoretical global solar radiation values were validated against 8,784 hourly averaged experimental data points, incorporating key solar parameters such as local time, solar declination, the Equation of Time (EoT), hour angle, and solar zenith angle. These theoretical inputs, combined with meteorological parameters (temperature, humidity, wind speed, and wind direction), were used to train multiple machine learning (ML) models. To improve model accuracy, a four-scenario data treatment strategy was applied, including Gaussian fitting and zero-removal techniques. In total, 28 ML models were evaluated, encompassing linear regression, support vector machines, Gaussian process regression, and neural networks. The Exponential GPR model emerged as the most accurate, achieving R2 values > 0.99 and RMSE as low as 0.13 W/m2. These results represent a significant performance improvement over existing models in the literature, many of which report RMSE values in the range of 20–60 W/m2 and R2 values between 0.85 and 0.98. The findings demonstrate substantial improvements over existing methods and offer a scalable solution for a more precise solar forecasting in various environments.http://www.sciencedirect.com/science/article/pii/S2215098625002113Gaussian Process RegressionMachine LearningMeteorological ParametersSolar RadiationTheoretical Modeling
spellingShingle Nader Ghareeb
Abeer Alanazi
Ahmad Sedaghat
Mohamad Hussein Farhat
Arash Mehdizadeh
Hayder Salem
Mohammad Nazififard
Ali Mostafaeipour
Integrating experimental and theoretical approaches for enhanced machine learning modeling of solar radiation
Engineering Science and Technology, an International Journal
Gaussian Process Regression
Machine Learning
Meteorological Parameters
Solar Radiation
Theoretical Modeling
title Integrating experimental and theoretical approaches for enhanced machine learning modeling of solar radiation
title_full Integrating experimental and theoretical approaches for enhanced machine learning modeling of solar radiation
title_fullStr Integrating experimental and theoretical approaches for enhanced machine learning modeling of solar radiation
title_full_unstemmed Integrating experimental and theoretical approaches for enhanced machine learning modeling of solar radiation
title_short Integrating experimental and theoretical approaches for enhanced machine learning modeling of solar radiation
title_sort integrating experimental and theoretical approaches for enhanced machine learning modeling of solar radiation
topic Gaussian Process Regression
Machine Learning
Meteorological Parameters
Solar Radiation
Theoretical Modeling
url http://www.sciencedirect.com/science/article/pii/S2215098625002113
work_keys_str_mv AT naderghareeb integratingexperimentalandtheoreticalapproachesforenhancedmachinelearningmodelingofsolarradiation
AT abeeralanazi integratingexperimentalandtheoreticalapproachesforenhancedmachinelearningmodelingofsolarradiation
AT ahmadsedaghat integratingexperimentalandtheoreticalapproachesforenhancedmachinelearningmodelingofsolarradiation
AT mohamadhusseinfarhat integratingexperimentalandtheoreticalapproachesforenhancedmachinelearningmodelingofsolarradiation
AT arashmehdizadeh integratingexperimentalandtheoreticalapproachesforenhancedmachinelearningmodelingofsolarradiation
AT haydersalem integratingexperimentalandtheoreticalapproachesforenhancedmachinelearningmodelingofsolarradiation
AT mohammadnazififard integratingexperimentalandtheoreticalapproachesforenhancedmachinelearningmodelingofsolarradiation
AT alimostafaeipour integratingexperimentalandtheoreticalapproachesforenhancedmachinelearningmodelingofsolarradiation