Manipulation of static and dynamic data center power responses to support grid operations

This research investigates three frameworks for data centers to deliver fast frequency response services from their UPS storage systems, HVAC cooling units, and the ability of off-gridding the entire data center during transient frequency events. The aim of this is to provide dynamic injection during transients in real time for grid operators and for power sellers in hour ahead markets. A static and a dynamic model was developed in DIgSILENT PowerFactory. In the static model, the data centers are off-gridded in response to emergency frequency signals from system operator in a centralized manner, whereas in the dynamic model the UPS systems and shiftable cooling units power consumption was altered continuously in the data centers in response to local frequency measurements considering different real-time scenarios. The performance of the proposed operational frameworks is validated, and calibrated to actual frequency events that occurred in the Irish power system. The sensitivity analysis demonstrates that both static and dynamic load controls can significantly improve system frequency metrics, i.e., frequency nadir and rate of change of frequency during transients. The key findings show that demand response can only make a substantial frequency improvement if a large amount of energy delivered within the timeframe of inertial response.