摘要
为提高数据中心空调运行性能,对某数据中心建立模型,对机柜出口温度在机箱不同的负载率下进行仿真抽样,建立CFD数据集。基于CFD数据集,分别使用三种机器学习模型即:Elman神经网络、RBF神经网络和SVR支持向量回归机,对数据中心稳态运行情况下的不同机柜负载率时空调回风温度进行快速预测。仿真结果表明:三种模型均在短时间内实现了较为准确的预测,但SVR支持向量回归机凭借其训练过程简洁、精度高、训练速度快等优势,更有效地预测热负荷温度。
To improve the performance of data centre air conditioning,a data centre is modeled and the cabinet outlet temperatures at different load rates of the chassis is simulation sampled to establish a CFD dataset.Based on the dataset,three machine learning models,i.e.,Elman neural network,RBF neural network and SVR support vector regression machine,are used to quickly predict the air conditioning return air temperature at different cabinet load rates under steady-state operation of the data centre.The simulation results show that all three models achieve more accurate prediction in a short time,but the SVR is more effective in predicting the heat load temperature by virtue of its advantages of concise training process,high accuracy and fast training speed.
作者
殷佳辉
朱兵
张一鸣
黄金森
苗益川
Yin Jiahui;Zhu Bing;Zhang Yiming;Huang Jinsen;Miao Yichuan(College of Electrical Engineering,Guizhou University,Guiyang,Guizhou 550025,China)
出处
《计算机时代》
2023年第11期71-75,78,共6页
Computer Era
基金
贵州省科技支撑计划项目(No.2017YFB0902100)。
关键词
数据中心
气流模拟
机器学习
快速预测
热参数
data center
airflow simulation
machine learning
fast prediction
thermal parameters
