Lingyun Xie , Kui Shan , Hong Tang , Shengwei Wang
{"title":"AI-empowered online control optimization for enhanced efficiency and robustness of building central cooling systems","authors":"Lingyun Xie , Kui Shan , Hong Tang , Shengwei Wang","doi":"10.1016/j.adapen.2025.100220","DOIUrl":null,"url":null,"abstract":"<div><div>Adopting Artificial Intelligence for optimizing building system controls has gained significant attention due to the growing emphasis on building energy efficiency. However, substantial gaps remain between academic research and the practical implementation of AI-based algorithms. Key factors hindering implementation include computational efficiency requirements and concerns about reliability in online applications. This paper addresses these challenges by presenting AI-empowered online control optimization technologies designed for practical implementation. A simplified deep learning-enabled Genetic Algorithm is developed to accelerate optimization processes, ensuring optimization intervals are short enough for online applications. This algorithm also significantly reduces CPU and memory usage, enabling deployment on miniaturized control station for field implementation. To enhance stability and reliability, a robust assurance scheme is introduced, which switches to expert knowledge-based control under abnormal conditions. Hardware-in-the-loop tests validate the proposed strategy's computation efficiency, control performance and operational robustness using a physical smart station controlling a simulated real-time dynamic cooling system. Test results show that the optimal control strategy achieves 7.66 % energy savings and exhibits strong operational robustness.</div></div>","PeriodicalId":34615,"journal":{"name":"Advances in Applied Energy","volume":"18 ","pages":"Article 100220"},"PeriodicalIF":13.0000,"publicationDate":"2025-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Advances in Applied Energy","FirstCategoryId":"1085","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S2666792425000149","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"ENERGY & FUELS","Score":null,"Total":0}
引用次数: 0
Abstract
Adopting Artificial Intelligence for optimizing building system controls has gained significant attention due to the growing emphasis on building energy efficiency. However, substantial gaps remain between academic research and the practical implementation of AI-based algorithms. Key factors hindering implementation include computational efficiency requirements and concerns about reliability in online applications. This paper addresses these challenges by presenting AI-empowered online control optimization technologies designed for practical implementation. A simplified deep learning-enabled Genetic Algorithm is developed to accelerate optimization processes, ensuring optimization intervals are short enough for online applications. This algorithm also significantly reduces CPU and memory usage, enabling deployment on miniaturized control station for field implementation. To enhance stability and reliability, a robust assurance scheme is introduced, which switches to expert knowledge-based control under abnormal conditions. Hardware-in-the-loop tests validate the proposed strategy's computation efficiency, control performance and operational robustness using a physical smart station controlling a simulated real-time dynamic cooling system. Test results show that the optimal control strategy achieves 7.66 % energy savings and exhibits strong operational robustness.