Search for: All records

Abstract A thermo-economic analysis (TEA) of a novel cooling and enhanced heat recovery (CEHR) system for data centers (DCs) is presented. Three financial metrics (net present value—NPV, return on investment—ROI, and payback period—PP) are calculated for hot and chilled water generation. Hot water generation uses vapor recompression to produce water at approximately 75 °C. Chilled water generation builds upon the hot water generation scenario by feeding the hot water stream into an absorption chiller. Without considering the additional costs for connecting the infrastructure with the customer, a payback period shorter than 2 years is found for a hot water generation system for a base case assuming a 10-MW data center (DC) in Philadelphia, PA, when carbon (reduction) credits are included. Chilled water generation is found to be economically unfavorable in this location. Sensitivities of economics to data center power, hot water versus chilled water generation, geographic region, and carbon credits are evaluated for five additional global locations in Europe and Asia. The economies of scale enable favorable payback periods for integrating hot water generation for facilities beyond 7 MW. Hot water generation is especially favorable in Singapore when replacing natural gas-based heating or hot water heat pumps. 

Abstract A significant number of investigations have been performed to develop and optimize cold plates for direct-to-chip cooling of processor packages. Many investigations have reported computational simulations using commercially available computational fluid dynamic tools that are compared to experimental data. Generally, the simulations and experimental data are in qualitative agreement but often not in quantitative agreement. Frequently, the experimental characterizations have high experimental uncertainty. In this study, extensive experimental evaluations are used to demonstrate the errors in experimental thermal measurements and the experimental artifacts during testing that lead to unacceptable inconsistency and uncertainty in the reported thermal resistance. By comparing experimental thermal data, such as the temperature at multiple positions on the processor lid, and using that data to extract a meaningful measure of thermal resistance, it is shown that the data uncertainty and inconsistency are primarily due to three factors: (1) inconsistency in the thermal boundary condition supplied by the thermal test vehicle (TTV) to the cold plate, (2) errors in the measurement and interpretation of the surface temperature of a solid surface, such as the heated lid surface, and (3) errors introduced by improper contact between cold plate and TTV. A standard thermal test vehicle (STTV) was engineered and used to provide reproducible thermal boundary conditions to the cold plate. An uncertainty analysis was performed in order to discriminate between the sources of inconsistencies in the reporting of thermal resistance, including parameters such as mechanical load distribution, methods for measuring the cold plate base, and TTV surface temperatures. A critical analysis of the classical thermal resistance definition was performed to emphasize its shortcomings for evaluating the performance of a cold plate. It is shown that the thermal resistance of cold plates based on heat exchanger theory better captures the physics of the heat transfer process when cold plates operate at high thermodynamic effectiveness. 

Abstract This study aims to improve the combined energy efficiency of data center cooling systems and heating/cooling systems in surrounding premises by implementing a modular cooling approach on a 42 U IT rack. The cooling solution uses a close-coupled technique where the servers are air-cooled, and the air in turn is cooled within the rack enclosure using an air-to-refrigerant heat exchanger. The refrigerant passively circulates in a loop as a thermosyphon, making the system self-sustaining during startup and shutdown, self-regulating under varying heat loads, and virtually maintenance-free by eliminating mechanical parts (other than the cabinet fans). A heat load range of 2 kW–7.5 kW is tested on a prototype system. Experimental results reveal stable thermosyphon operation using R1233zd(E) as the working fluid, a maximum evaporator pressure drop of 21.5 kPa at the highest heat load and a minimum thermosyphon resistance of 6.8 mK/W at a heat load of 5.7 kW. The air temperature profile across the load banks (server simulators) and evaporator follow the same profiles with varying heat loads. Heat losses from the cabinet due to natural convection and radiation are of the order of several Watts for heat loads below 4 kW and rise sharply to 1 kW at the highest heat load tested. The system time constant is determined to be 25 min. The heat recovery process can be financially and environmentally beneficial depending on the downstream application. 

Abstract A thermosyphon-based modular cooling approach offers an energy efficient cooling solution with an increased potential for waste heat recovery. Central to the cooling system is an air-refrigerant finned tube heat exchanger (HX), where air is cooled by evaporating refrigerant. This work builds on a previously published two-dimensional (2D) model for the finned-tube HX by updating and validating the model using in-house experimental data collected from the proposed system using R1233zd(E) as the working fluid. The results show that key system variables such as refrigerant outlet quality, air and refrigerant outlet temperatures, and exchanger duty agree within 20% of their experimental counterparts. The validated model is then used to predict the mean heat transfer coefficient on the refrigerant side for each tube in the direction of airflow, indicating a maximum heat transfer coefficient of nearly 1200 W/(m2 K) for a HX duty of 5.3 kW among the tested cases. The validated model therefore enables accurate predictions of HX performance and provides insights into improving the heat exchange efficiency and the corresponding system performance. 

Abstract The energy demands from data centers contribute greatly to water scarcity footprint and carbon emissions. Understanding the use of on-site renewable power generation is an important step to gain insight into making data centers more sustainable. This novel study examines the impact of on-site solar or wind energy on data center water scarcity usage effectiveness (WSUE) and carbon usage effectiveness (CUE) at a U.S. county scale for a given data center size, water consumption level, and energy efficiency. The analysis uncovers combinations of specific metrics associated with grid-based carbon emissions and water scarcity footprint that enable predictions of the improvements anticipated when implementing on-site solar or wind energy. The implementation of on-site renewables has the most benefit in reducing carbon footprint in areas with high existing grid-based emissions such as the western side of the Appalachian Mountains (e.g., central and eastern Kentucky). The largest benefit in reducing water scarcity footprint is generally seen in counties with low water scarcity compared to adjacent areas (e.g., northern California).