基于数据挖掘方法的数据中心能耗与空调系统关键参数量化关系研究
Quantitative relation between energy consumption and key air conditioning system parameters in data centers based on data mining method
摘要：
使用Lasso回归筛选和人工参数筛选2种方式确定了数据中心能耗预测模型的输入参数，利用XGBoost算法对北京某数据中心的能耗和空调系统数据进行了数据挖掘，实现了对该数据中心电能利用效率（PUE）的准确预测，并得到了空调系统参数对PUE的定量化影响相关性排序。验证了使用数据挖掘方法筛选预测模型参数的可靠性。
关键词：数据中心,空调系统,电能利用效率（PUE）,能耗,数据挖掘,影响因素
Abstract：
Uses Lasso regression screening and manual parameter screening to filtrate the input parameters of the energy consumption prediction model of data centers. Implements the XGBoost algorithm to do data mining about the energy consumption and air conditioning system parameters of a data center in Beijing. Achieves the accurate prediction of the power usage effectiveness (PUE) of the data center. Observes the quantitative influence correlation rank of the air conditioning system parameters on the PUE. Verifies the reliability of the parameter selection by the data mining method.
Keywords:data center, air conditioning system, power usage effectiveness (PUE), energy consumption, data mining, influence factor