Demand Response for Reducing Coincident Peak Loads in Data Centers
- Maciej Lukawski, Cornell University, Ithaca, New York, United States
- Jefferson W. Tester, Chemical and Biomolecular Engineering, Cornell University, Ithaca, New York, United States
- Michal C. Moore, Cornell University, Ithaca, New York, United States
- Pawel Krol, AGH University of Science and Technology, Kraków, Poland
- C. Lindsay Anderson, Cornell University, Ithaca, New York, United States
AbstractDemand response is a key aspect of managing uncertainty and reducing peak loads in electric grids. This paper considers the capability of a datacenter to provide responsiveness to grid signals through cooling system control. The strategy is based on pre-cooling the center for provision of load reduction during demand response events, and is evaluated using a numerical model of a cooling system, validated against experimental data obtained from a small telecommunication data center. The pre-cooling strategy is applicable to a wide-range of demand response programs, but is illustrated on the example of an established critical peak pricing program; specifically the 4 coincident peak (4CP) program in the ERCOT ISO. Precooling reduced the annual cost of electricity used by the cooling system by 7.8% to 8.6%, while increasing the total energy use only by 0.05%. This translated into 2% to 2.6% reduction in the electric bill of the whole data center. The developed demand response strategy is suitable for data centers with power densities below 500 W/m2 which do not use server air containment systems.
Return to previous page