Compact models for real-time modeling and control of data centers

Data centers are massive facilities that house large numbers of computer servers and auxiliary equipment. Today, the increasing heat loads and energy costs associated with the cooling of data centers are driving a substantial research effort to develop more energy efficient means of data center thermal management that will require the use of reduced-order or compact models. The focus of this dissertation is the development of a complement of compact modeling approaches that run in real-time for use in the control of data center thermal management systems. The first approach utilizes a computational fluid dynamics (CFD) simulation-based Artificial Neural Network (ANN) model that was applied to a basic hot aisle/cold aisle raised floor data center configuration to predict various data center operating conditions. The second approach involved the use of a partitioning methodology where a characteristic configuration was used to analyze a more complex room configuration. Specifically, a configuration of aligned and anti-aligned cooling units was analyzed using CFD simulations for a multiple hot aisle/cold aisle room in order to partition the room layout into a finite number of characteristic zones where the ANN models readily apply. The zonal modeling approach used commonly to analyze building ventilation and energy transport was applied to basic data center configurations. The predictions from the zonal modeling were compared to the results from full CFD simulations. Finally, the other reduced-order models that were investigated to enable real-time assessment of the operating and control conditions required to improve data center energy efficiency were the Proper Orthogonal Decomposition (POD) based methods. The real benefit of the compact models is that they can be coupled with optimization algorithms and integrated into real-time control systems of data center cooling systems in order to achieve significant energy savings for a wide variety of configurations and operating conditions in both existing and newly designed data centers.