计算机应用研究2026，Vol.43Issue(8)：2250-2260,11.DOI:10.19734/j.issn.1001-3695.2025.12.0523
数据中心HVAC智能控制的研究现状与挑战
Research status and challenges of intelligent HVAC control in data centers
摘要
Abstract
With data centers scaling up and low-carbon operation requirements rising,cooling systems must balance thermal safety and PUE.High heat flux,heterogeneous cooling chains,and operating uncertainty make conventional rule-based control difficult to achieve both efficiency and adaptability.To clarify the state of the art and support engineering deployment,this re-view surveyed intelligent HVAC control for data centers,focusing on paradigm taxonomy,key-feature comparison,and coordi-nated control scopes.It categorized prior studies by control mechanism into rule-based control,model predictive control,on-line reinforcement learning,and offline reinforcement learning,and compared them in terms of interpretability,constraint han-dling,global optimization capability,data requirements,and deployment safety mechanisms.It further summarized hierarchi-cal coordination across air-cooled room terminals,chiller plants,economizers,and liquid-cooling loops,and discussed bottle-necks in model mismatch,prediction uncertainty,policy generalization,and robust,reproducible empirical evaluation,along with future directions.关键词
暖通空调/基于规则的控制/模型预测控制/强化学习/离线强化学习Key words
HVAC/rule-based control/model predictive control/reinforcement learning/offline reinforcement learning分类
信息技术与安全科学引用本文复制引用
王浩楠,万剑雄,李进,刘楚仪,李雷孝..数据中心HVAC智能控制的研究现状与挑战[J].计算机应用研究,2026,43(8):2250-2260,11.基金项目
内蒙古自治区2025年首批"五大任务"关键技术研究专项(NMGW-DRW2025-03) （NMGW-DRW2025-03）
内蒙古自治区科技计划资助项目(2025KYPT0036,2025YFHH0213,2025QN06016,2025SYFHH1239,2026MS0303,2025KYPT0014) （2025KYPT0036,2025YFHH0213,2025QN06016,2025SYFHH1239,2026MS0303,2025KYPT0014）
一流学科科研专项项目(YLXKZX-NGD-026) （YLXKZX-NGD-026）
内蒙古自治区高等学校创新团队发展计划基金资助项目(NMGIRT2506) （NMGIRT2506）
