- 上一篇：氟伐他汀抑制昆虫保幼激素合成英文文献和中文翻译
- 下一篇：注塑模具的设计与热分析英文文献和中文翻译
HVAC systems are the dominant component of non-ITenergy demand, owing to the large amount of heat generated bythe IT equipment that must be removed from the interior space.HVAC energy use can be affected by local climate conditions,especially when incorporating economizers to cool the IT equip-ment by directly routing outside air into the data center duringfavorable weather conditions. Building location also affects green-house-gas emissions associated with data center operation, owingto regional differences in the mix of primary energy used forgenerating electricity.This paper builds on previous data center modeling efforts byaddressing climate and mechanical equipment differences amongdata center types and evaluating the consequences of thesedifferences for energy use. Cities with significant data centeractivity are identified. The climate conditions of these cities areapplied to data center energy models that have been tailored torepresent different data center space types. The results of thisanalysis quantitatively demonstrate how building location andeconomizer use can influence energy demand and greenhouse-gasemissions. Aspects of data center operation are highlighted wherethe potential for improvement is large.2. MethodsEnergy use associated with data center operation in the UnitedStates is estimated using this equation:Etotal¼s;rITs;r PUEs;r (1)where Etotal represents the energy required for all data centeroperations in the US and ITs,r is the operational energy use associ-ated with IT equipment for space type s in region r. Total data centerbuilding energy use is calculated as a function of the IT energythrough the PUE metric, where PUEs,r represents the estimatedannual average energy use performance of the non-IT equipmentassociated with data center space type s in climate region r,expressed as the ratio of total energy use to IT energy use in a datacenter building.For the purposes of this paper, the space type-specificITequipment energy estimates presented in Table 1 are equallydistributed into five representative US cities so as to explore theeffect of climate differences on building operation among prom-inent data center locations. The five cities e San Francisco (CA), Seattle (WA), Chicago (IL), Dallas (TX), and Richmond (VA) e wereselected after analysis of two data sets that are presented in theSupplementary Material section of this article, available on-line atthe journal’s website. First, commercial buildings with significantdata center activity were identified from
Commercial BuildingEnergy Consumption Survey (CBECS) data [10]. Installed serversdocumented in that data set were disaggregated into censusregions. Second, a list of US metropolitan areas with largeconcentrations of existing data centers previously compiled fromUS Department of Energy data [1] was used to create a list ofspecific cities with significant data center activity. The cities fromthis list were then matched to the corresponding census region.Since a more refined US distribution of data center activity is notcurrently available,we base the energy analysis reported here on anassumed equal baseline distribution of national data center energyuse among the five cities. Each of these cities has significant datacenter activity and they collectively span much of the range of USclimatic conditions.Operational non-IT data center energywas estimated, specifictobuilding size and location, using a custom-built analytical model.The same energy modeling approach has been used in previousstudies [11e13] and is based on a combination of fundamentalHVAC sizing equations and equipment characteristics observedthrough professional experience. A custom-built energy model wasused since conventional energy modeling programs (e.g., DOE-2)are not designed to incorporate some of the HVAC characteristicsunique to data centers, such as high return air temperatures(>22 C) and high internal load densities. Data centers have floor-area-weighted power densities that are 15e100 times as large asthose of typical commercial buildings [14]. In this study, the heatgenerated from data center occupants and heat transfer throughthe building envelope were assumed to be negligible relative to theheat produced by IT equipment.As outlined in Table 2, energy use estimates were modeled fora reference data center design (Baseline) that provides minimalventilation air and represents conventional data center operation.Such a design has a high intensity of energy demand owing to theexclusive use of compressor-based cooling to remove internal heatloads. Two HVAC economizer designs aremodeled: Economizer andEconomizer Plus. These economizer designs use air-side econo-mizers to supply large flow rates of outside air into the data centerduring cool weather conditions. (An alternative design that was notanalyzed in this study is thewater-side economizer,which employscooling towers to provide chilled water in the cooling system.Water-side economizers avoid exposure of IT equipment to excessoutside air, but the energy savings potential of these systems arelimited in many climates.) In practice, temperature and humiditycontrols are used to determinewhen the economizers are active. Inthe present study, airflow in the economizer designs is identical tothe baseline design during periods when the economizer is inac-tive. The economizers operate when both the outside air temper-ature is less than the return air temperature setpoint and theoutside air dewpoint is less than the return air drybulb temperatureat the maximum allowed relative humidity (RH) in the space (i.e.,Table 2Modeled operational settings for each design scenario.Baseline Economizer ScenariosEconomizer Economizer PlusEconomizer use No Yes YesSupply/return temperature ( C) 18/22 18/22 18/29Humidity restrictions (RH) 40e55% 40e55% 1e100%Upper drybulb lockout ( C) n/a 22 29Upper dewpoint lockout ( C) n/a 19 29 the upper dewpoint economizer lockout). The Economizer simula-tions represent data centers using economizers while maintainingconventional indoor temperature and humidity settings [15]. Thesesame settings are applied in the Baseline case.