Fast Frequency Response From Data Centres, Using Machine Learning

The current research work develops a system wide fleet of data centres using data-driven approach. The models are then integrated with power system models, including a unit commitment and economic dispatch model and a stability analysis tool. Subsequently, the impacts of demand response on power system scheduling and dynamic behaviour are investigated, based on a number of flexible load control strategies. It is shown that providing a fast frequency response (FFR) from data centres improves the system frequency profile, with minimum disruption to end user comfort. The form of the reserve activation technique, and its speed of activation, is also shown to be important in determining the shape of the frequency response. The all-island power system of Northern Ireland and Republic of Ireland, with a horizon of 2030, is considered as a representative study system.