基于SVR数据中心空调系统瞬态热参数预测

Transient thermal parameter prediction based on SVR data center air conditioning system

数据中心空调系统是维持数据中心的关键设备,直接影响到数据中心的安全运行,目前对空调系统的研究大多集中在节能降耗以及气流优化等领域。在空调故障影响等瞬态变化领域的研究仍然较少。因此有必要探究空调系统故障对机房气流组织的影响,建立针对空调失效极端工况下的快速温度预测模型,为能效控制系统及运行系统提供参考。本文根据空调系统故障实验分别建立了空调冷冻水泵失效及风机失效情况下的关键位置的温度变化时间序列预测模型,模型基于线性核函数支持向量回归机。研究表明相较于非线性核函数支持向量机,线性核函数支持向量机更适合进行冷冻水泵失效时的热参数预测。

The data center air conditioning system is a key equipment to maintain the data center,which directly affects the safe operation of the data center.Most of the current research on air conditioning system is focused on energy saving and consumption reduction as well as airflow optimization.Fewer studies have emerged in the area of transient changes such as the impact of air conditioning failures.Therefore,it is necessary to explore the impact of air conditioning system failure on the airflow organization of the server room and establish a fast temperature prediction model for the extreme operating conditions of air conditioning failure to provide a reference for energy-efficient control systems and operation systems.In this paper,time series prediction models of temperature changes at critical locations under air conditioner chilled water pump failure and fan failure are established based on linear kernel function support vector regression machine.The research demonstrates that the linear kernel function support vector machine is more suitable for the prediction of thermal parameters in case of chilled water pump failure than the nonlinear kernel function support vector machine.