数据中心高密度机柜的风冷系统及其散热能力分析OA
Analysis of Air-Cooling System and Heat-Dissipation Capacity for High-Density Cabinets in Data Centers
随着信息技术的飞速发展和智能算力的广泛应用,单机柜的功率密度不断攀升.这种高功率密度配置在显著提升计算能力的同时,也带来了严峻的散热挑战.针对当前仍以风冷技术为主的数据中心,本文对微通道两相自驱动机柜级风冷系统进行了实验测试,测试结果表明:该系统可实现40 kW的换热能力.将该系统应用于太原某高密度机柜的示范工程中,实测数据显示,在满载条件下,机柜发热功率达到30 kW,本系统能够有效满足其散热需求.并基于实验数据构建了换热过程模型,研究了未来两相自驱机柜级风冷末端在不同外部条件下的换热能力,以期探索风冷系统的换热能力极限.结果显示:在机柜级末端回风温度维持在35℃、冷水源温度低至12℃、末端迎面风速高达5 m/s的条件下,风冷末端可实现65.2 kW的散热量.通过实验验证和理论分析,可为存量风冷数据中心进一步提升算力提供新的可能性和技术支撑.
With the rapid development of information technology and the wide application of intelligent arithmetic,the power density of a single cabinet continues to increase.Such high-power-density configurations cause severe cooling challenges while significantly increasing computing power.This study tests a two-phase microchannel self-driven cabinet-level air-cooling system for data centers dominated by air-cooling technology.The test results show that the system can achieve a heat-transfer capacity of 40 kW.The system is subsequently applied to a high-density cabinet demonstration project in Taiyuan.The measured data show that when the power of the cabinet reaches 30 kW under full-load conditions,this system can effectively satisfy its heat-dissipation requirements.In addition,this study constructs a heat-transfer model based on experimental data and further explores the heat-transfer capability of two-phase self-driven cabinet-level air-cooled terminals under different external conditions,with the aim of exploring the limits of air-cooling systems.The results show that the air-cooled terminal can realize a heat dissipation of 65.2 kW if the return air temperature of the cabinet-level terminal is maintained at 35℃,the temperature of the chilled water source is as low as 12℃,and the wind speed on the terminal reaches 5 m/s.Based on experimental validation and theoretical analysis,this study offers new possibilities and technical support for stock air-cooled data centers to further improve their arithmetic power.
王馨翊;王露;陈晓轩;李弘锐;李震
清华大学航天航空学院 北京 100084清华大学航天航空学院 北京 100084清华大学航天航空学院 北京 100084清华大学航天航空学院 北京 100084清华大学航天航空学院 北京 100084
通用工业技术
智算高密度机柜风冷系统冷却能力
intelligent computinghigh-density cabinetair-cooling systemcooling capacity
《制冷学报》 2026 (1)
51-58,8
本文受河北省院士合作重点单位项目(225A9912D)资助.(The project was supported by the Hebei Province Academician Cooperation Key Unit Project (No. 225A9912D). )
评论