Data Center Journal

Volume 28 | August 2013

Issue link:

Contents of this Issue


Page 22 of 36

DESIGN CORNER BY DANA ETHERINGTON, CEM, LEED AP BD+C Simplified Data Center Energy Modeling Hourly Weather Analysis Coupled with Low-Cost CFD Modeling GETTING STARTED The analysis methodology outlined herein is intended to guide early phase decision making. This allows us to review high level energy saving strategies, seek rebates through local utility programs and identify 'white space' design issues early on. White space issues can range from by-pass/re-circulation airflow to isolated thermal "hot spots" and dangerous rack inlet temperatures. It is important to understand that the methodology described herein is not intended as a detailed design tool or to predict exact energy usage. The purpose is to develop reasonable baselines to which we can make comparisons. First we need to locate weather data for the specific site or a nearby weather station, whichever is available. Hourly weather data is available through the extensive solar database developed by the National Renewable Energy Laboratory (NREL). Typical Meteorological Year (TMY) Solar Data can be downloaded directly in editable spreadsheet format [1]. Year by year data is also available for select weather stations for the years 1991 to 2010 [2]. Next we need to identify which pieces of equipment we are going to analyze; chillers, cooling towers, pumps, CRAC units (DX or CHW), etc. Performance data should be obtained for all appropriate equipment. A simple example of this type of data is kW/ton efficiency ratings over a given capacity range for chillers. Control strategies for pumps and fans should also be identified (constant / variable speed). Once this information has been compiled, the analyst can begin filling in the spreadsheet. The hourly (8,760) spreadsheet will be referred to as the data center energy model from now on. The logical order of components within the energy model is month / day / time, then weather data and the central plant cooling load followed by energy consumption of equipment. An example of the chiller plant energy usage calculation is shown in Figure 1 below. The part-load efficiency of the chiller is in Column Q, modeled chiller energy consumption in Column R and the auxiliary component energy usage in the remaining columns. 20 | THE DATA CENTER JOURNAL Figure 1 - A "snapshot" of the 8,760 Hour Energy Model Having an hourly energy model allows the analyst to apply variations in IT loading and building cooling loads based on time of day and seasons. This is most readily accomplished through simple 24 hours schedules based on a percent of maximum. For example; the IT load could be modeled as 50% maximum from 8pm to 8am, and 100% at all other times for 365 days per year. The actual organization of the energy model is up to the individual analyst. Before any CFD modeling is carried out, it is important to choose the appropriate software package and modeler. Software packages vary in capabilities, required modeling time, simulation time, and of course licensing costs. These items must be reviewed in detail before selecting a particular software package. A novice CFD analyst may be capable of modeling a less complicated data center, whereas an experiences analyst should be identified to perform modeling of a more complex design. As with any modeling, the results are only as accurate the inputs utilized to derive them. Although software has made many advances in recent years, it is still up to the analyst to ensure the modeling is done accurately.

Articles in this issue

Links on this page

Archives of this issue

view archives of Data Center Journal - Volume 28 | August 2013