Optimization of row-and-rack level airflow management in data centers

Building energy use is responsible for around 40% of total primary energy use, and about 36% of total global carbon emission, causing significant environmental impacts (e.g., climate change). Datacenters (DCs) belong to the special building sector accounting for a very small proportion of the building volume but have a very high energy use. Electricity used for DCs accounts for about 2% of the global total electricity use, whose figure is predicted to reach 5% by 2024. Thus, energy conservation methods should be applied in DCs under the premise of safe operation. This thesis adopts some research methods (e.g., literature review, experimental tests, numerical studies,model validation, and formal analysis). A literature review can engender new directions and innovations and is the ground for future research and thesis. Thus, this thesis firstly conducts a literature review about the cooling system and optimization in DCs. Considering the current huge market share of air-side cooling and its high cooling reliability & applicability and normalized model matching design compared with liquid-side cooling, this thesis proposes row-and-rack level airflow management is worth further studying. Then, 4 different row-and-rack level methods are proposed and studied to optimize the thermal environment through experimental & numerical studies and model validation. The numerical models are validated by the experimental results,and the results of both experimental and numerical studies are presented and analyzed as well as energy saving potentials. Finally, the literature review is conducted again based on phase change cooling (PCC) in DCs, and PCC is recommended for future study. The results show that the numerical results are in good agreement with the experimental results, and the reliability and feasibility of numerical models are validated. In addition, all the row-and-rack level airflow management methods can improve the thermal environment in DCs to varying degrees. The rack hotspot temperatures can be decreased by 1.5-2.5 K with different methods, which reduces the risk of servers' down-time and extends their lifespan. Furthermore, some of these methods can achieve considerable energy savings (98-146 kWh electricity use per day) under the premise of safe operation. The proposed methods (server terminal baffles and tilted server placement) can be used directly in the operation phase of DCs, and only need simple modifications (e.g., adding baffles and adjusting server angles) for safe operation and thermal environment improvement, while the rest (in-rack UFAD and step-like server placement) are applicable in the DC design phase, which also just needs simple modifications (e.g., changing the positions of perforated floor and server rack, and rack length)