A dynamic model of failure scenarios of the dry cooler in a liquid cooled chiller-less data center

Data centers take up a significant portion of the national energy consumption in the U.S., accounting for approximately 2% of the total electricity used in 2010. Around the world that percentage is growing at a rapid rate. Within data centers, cooling systems constitute a large portion of the energy consumed, accounting for roughly 25-35% of total expenditures. Many new data center cooling technologies have been developed to improve cooling energy efficiency. A data center cooling facility proposed by IBM was constructed to reduce cooling energy use to less than 5% of the total Information Technology (IT) energy use, which was accomplished through a combination of warm water cooling servers and liquid-side economization. The heat load of IT equipment is rejected out to the ambient air without using a refrigeration/chiller plant or a Computer Room Air Handlers unit (CRAH), which consume a large amount of energy. This work develops simulation tools for the dry cooler, which is an air-to-liquid cross flow heat exchanger used in this cooling technology. Steady state test data is incorporated into the models, and is then used to validate them. The dynamic prediction accuracy of the simulation tool is also compared with real time measurement results of the dry cooler. Results of the simulation and testing are then used to obtain a more complete understanding of the dry cooler performance data and analyze the performance of the dry cooler in the cooling infrastructure. Failure analyses of the dry cooler are also performed in this work. One failure category of the dry cooler is parametrically modeled. The dynamic effects of failures of the dry cooler are reported and analyzed. The effects of the thermal mass of liquid and ambient air as well as the effects of natural convection and thermal, hydraulic characteristics of the external loop on the dynamic behavior of the dry cooler in this failure scenario are presented.