Data centre building energy optimization with differential temperature control utilizing I.T. power measurement

Due to increasing demand for artificial intelligence, data centre buildings could consume more than 20 % of utility electrical energy for some countries by 2030. Whilst research has been performed into energy reduction via data centre supply air temperature reset control, air containment and economy cycle free cooling, there is a significant research gap regarding differential temperature control via I.T. kW load measurement to improve economy cycle energy savings. This paper presents over 5 years of research for a new cooling system control method for co-location data centres and a new theoretical model with prototype site research validation testing, and development to assess energy efficiency improvement. The study investigated the reduction in energy consumption achieved by increasing differential and return air temperatures, increasing free-cooling economization, decreasing air flow rates, and resetting compute hot aisle temperatures when un-occupied. The research validated a new theoretical model developed with Design Builder™ and EnergyPlus™ through experimental testing of a working prototype to estimate energy use, Power Usage Effectiveness (PUE), and carbon emissions. Research was performed at two state-of-the-art facilities in Brisbane and Sydney, Australia, with I.T. design capacities of 6000 kW and 30,000 kW. The research conservatively estimated the new method could reduce data centre energy consumption of the central plant equipment by up to 10 % when the building load achieves 50 % of design maximum capacity. Greater savings could be achieved in the future as allowable supply and differential temperatures of the I.T. equipment increase to improve sustainability. The contributions include research, development, and modelling of a new and novel data centre cooling methodology utilizing differential temperature control based on IT power measurement, first of kind research validation of the new model and tuning of the new system to optimize the key performance metric PUE for data centres.