数据中心HVAC智能控制的研究现状与挑战OA

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.