Cooling Load Prediction via Support Vector Regression in Individual and Hybrid Approaches

Optimizing energy efficiency and minimizing environmental impact in buildings depends critically on managing cooling requirements. The application of Support Vector Regression Model to the prediction of cooling load is explored in this study. It enhances these models with two cutting-edge optimizati...

Full description

Saved in:
Bibliographic Details
Main Authors: Honglei Yao, Andrew Topper
Format: Article
Language:English
Published: Bilijipub publisher 2024-03-01
Series:Journal of Artificial Intelligence and System Modelling
Subjects:
Online Access:https://jaism.bilijipub.com/article_193318_57be23d6132acad9c046072f1d73b4d2.pdf
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849404811002249216
author Honglei Yao
Andrew Topper
author_facet Honglei Yao
Andrew Topper
author_sort Honglei Yao
collection DOAJ
description Optimizing energy efficiency and minimizing environmental impact in buildings depends critically on managing cooling requirements. The application of Support Vector Regression Model to the prediction of cooling load is explored in this study. It enhances these models with two cutting-edge optimization methods: Crystal Structure Algorithm and Reptile Search Algorithm. SVR, a machine learning technique renowned for its flexibility and interpretability, is used in this study to capture complex relationships between various building parameters and cooling load. The dependent variable, CL, is analyzed about various factors, including relative compactness, wall area, roof area, orientation, surface area, overall height, glazing area, and glazing area distribution. SVR proves to be adept at understanding non-linear relationships, making it suitable for these kinds of applications. The modelling process gains intelligence from the combination of RSA and CSA optimizers. Building management systems stand to benefit greatly from this advancement, which will allow for more precise control over cooling systems and efficient use of energy. Moreover, the hybrid SVCS model, with its minimal RMSE value of 0.747 and remarkable R2 value of 0.994, consistently yields reliable results for CL prediction. This study advances the field of energy-efficient building management by demonstrating how machine learning methods and clever optimization algorithms can be used to predict cooling loads accurately.
format Article
id doaj-art-2c8f6a7dddda4ac8a12074fd570dbbef
institution Kabale University
issn 3041-850X
language English
publishDate 2024-03-01
publisher Bilijipub publisher
record_format Article
series Journal of Artificial Intelligence and System Modelling
spelling doaj-art-2c8f6a7dddda4ac8a12074fd570dbbef2025-08-20T03:36:53ZengBilijipub publisherJournal of Artificial Intelligence and System Modelling3041-850X2024-03-010201668210.22034/jaism.2024.445961.1027193318Cooling Load Prediction via Support Vector Regression in Individual and Hybrid ApproachesHonglei Yao0Andrew Topper1State Grid Shandong Electric Power Research Institute, ChinaUniversity of Grand Valley State, Grand Rapids, Michigan, 49504, United StatesOptimizing energy efficiency and minimizing environmental impact in buildings depends critically on managing cooling requirements. The application of Support Vector Regression Model to the prediction of cooling load is explored in this study. It enhances these models with two cutting-edge optimization methods: Crystal Structure Algorithm and Reptile Search Algorithm. SVR, a machine learning technique renowned for its flexibility and interpretability, is used in this study to capture complex relationships between various building parameters and cooling load. The dependent variable, CL, is analyzed about various factors, including relative compactness, wall area, roof area, orientation, surface area, overall height, glazing area, and glazing area distribution. SVR proves to be adept at understanding non-linear relationships, making it suitable for these kinds of applications. The modelling process gains intelligence from the combination of RSA and CSA optimizers. Building management systems stand to benefit greatly from this advancement, which will allow for more precise control over cooling systems and efficient use of energy. Moreover, the hybrid SVCS model, with its minimal RMSE value of 0.747 and remarkable R2 value of 0.994, consistently yields reliable results for CL prediction. This study advances the field of energy-efficient building management by demonstrating how machine learning methods and clever optimization algorithms can be used to predict cooling loads accurately.https://jaism.bilijipub.com/article_193318_57be23d6132acad9c046072f1d73b4d2.pdfbuilding energycooling loadsupport vector regressionreptile search algorithmcrystal structure algorithm
spellingShingle Honglei Yao
Andrew Topper
Cooling Load Prediction via Support Vector Regression in Individual and Hybrid Approaches
Journal of Artificial Intelligence and System Modelling
building energy
cooling load
support vector regression
reptile search algorithm
crystal structure algorithm
title Cooling Load Prediction via Support Vector Regression in Individual and Hybrid Approaches
title_full Cooling Load Prediction via Support Vector Regression in Individual and Hybrid Approaches
title_fullStr Cooling Load Prediction via Support Vector Regression in Individual and Hybrid Approaches
title_full_unstemmed Cooling Load Prediction via Support Vector Regression in Individual and Hybrid Approaches
title_short Cooling Load Prediction via Support Vector Regression in Individual and Hybrid Approaches
title_sort cooling load prediction via support vector regression in individual and hybrid approaches
topic building energy
cooling load
support vector regression
reptile search algorithm
crystal structure algorithm
url https://jaism.bilijipub.com/article_193318_57be23d6132acad9c046072f1d73b4d2.pdf
work_keys_str_mv AT hongleiyao coolingloadpredictionviasupportvectorregressioninindividualandhybridapproaches
AT andrewtopper coolingloadpredictionviasupportvectorregressioninindividualandhybridapproaches