ScienceDirect
w w w . i i fi i r . o r g
Article history: This study introduced a novel energy saving cooling system, i.e. a combined cycle coupled
Received 1 August 2014 with a traditional vapor compression cycle with a pumped liquid two-phase cooling cycle.
Received in revised form The system has two operation modes, i.e. the compression cycle mode driven by
18 September 2014 compressor and the pump cycle mode driven by refrigerant pump. A multi-purpose test
Accepted 28 September 2014 bench was constructed to experimentally evaluate the performance of the integrated cycle
Available online 7 October 2014 system under various operation conditions. The effects of cycle working condition and the
shift temperature between the two operation modes on the overall cycle performance were
Keywords: investigated in detail. It is found that the novel cycle system has a higher EER compared to
Data center the traditional compressor system when the ambient temperature is relatively low. The
Liquid refrigerant pump further experimental results and comparative annual energy saving analysis also indicated
Cooling that the proper shift temperature is about 5 C from the system EER and cooling capacity
EER point of view.
Shift temperature © 2014 Elsevier Ltd and IIR. All rights reserved.
R22
where a cool air climate is available during most of the year, the vapor compressor cycle system. Such a novel cycle system
and two-phase cooling reduced thermal resistance as much as may be more effective for improving energy efficiency and
an order of magnitude less than that of air and significantly cost saving of data centers. The performance of the novel
below that of liquid cooling. Additionally, the thermal resis- cooling system driven by liquid refrigerant pump and
tance of the two-phase cooling by the force-fed manifold compressor in data centers is experimentally evaluated.
microchannels is 20 times lower than that of the liquid cooling
at 5 times lower pumping power consumption. The phase
change cooling systems will have significant competitive
advantage over conventional systems due to the lower costs 2. Experimental setup
and energy savings. Marcinichen et al. (2010, 2012) and Wu
et al. (2013) proposed three two-phase cooling cycles for Fig. 1 illustrates the integrated cooling cycle driven by a liquid
cooling data center servers and the cycles using two-phase refrigerant pump and a vapor compressor and it can be
boiling in microchannels for removing the heat from the mi- operated in two different operation modes by means of the
croprocessors and memories. The results showed that the solenoid valves (SVs) 1 to 4. The choice of the system opera-
pumping power consumption is on the order of 5 times higher tion mode depends on the cooling demand and ambient air
for the water-cooled cycle when considering the potential temperature. When the ambient air temperature is low, the
saving by implementing on-chip cooling with waste heat re- system is shifted to the pump operation mode by means of
covery. The results also showed that, when compared with opening the SVs 1 and 3, and shutting off SVs 2 and 4. In the
the traditional air-cooling systems, the energy consumption liquid pump operation cycle, a feed custom gear pump with
of the data center could be reduced by 50% when using a liquid constant speed compresses the working fluid from the satu-
pumping cycle and by 41% when using a vapor compression rated liquid state at the receiver to a subcooled liquid state at
cycle. Further investigation indicated that a PI structure and pump outlet (process 5e6). Subsequently, the subcooled fluids
simple SISO and SIMO strategies were sufficient to fulfill the absorb the data center heat in the evaporator and then change
requirements. to a saturated or superheat vapor state (process 8e1). The high
However, previous studies of two-phase cooling in data pressure and temperature vapor enters the connected pipe
centers mainly focused on the application in a microchannels between the evaporator and condenser, and then down to a
evaporator for removing the heat of on-chip cooling. In a lower pressure and temperature (process 1e2e3). Lastly, the
pump liquid two-phase cycle, the potential of energy saving heat will be rejected to the ambient in the condenser where
with lower input power feature has been already exploited, working fluid changes from the superheat state to the liquid
where it is the basis of improved cycle performance. In fact, state at the inlet of the receiver (process 3e4). The major
the two-phase cooling can be used to modify the existing function of the check valves (CVs) 1 to3 in the loop diagram is
tradition air-cooled systems in data centers considering the to ensure the proper flow direction of working fluids in
energy and cost savings, i.e. obtaining a higher system per- different operation modes. When designing a pumping sys-
formance with a lower pump input power during lower tem, it is important to prevent the phenomenon of cavitation.
ambient temperatures. For this purpose, this paper presents a Cavitation occurs when bubbles form in the liquid at the
novel hybrid cooling cycle coupled with liquid refrigerant pump inlet. In order to make the liquid pump work stable and
pump cycle and vapor compressor cycle to be used in data avoid cavitation, the 6 L receiver is lifted up to 1.5 m higher
centers. Based on existing vapor compressor cycle, a liquid than the liquid pump. The liquid refrigerant pump cycle is
refrigerant pump associated with a receiver is introduced into characterized with low condensing temperature and high
overall efficiency. This is a good operation option during the as indicated in Fig. 1. The flow rate of the working fluid was
winter season. measured using a turbine-type flow meter with an accuracy
When the ambient temperature is high, the system is of ±0.5%. One sight glass was installed downstream of the
shifted to the compressor operation mode by means of Micro Motion mass flow meter to identify the case of two-
opening SVs 2 and 4 and shutting off SVs 1 and 3. In the phase flow. The electric power output of the generator (i.e.
compressor operation cycle, a VR30KM-TFP type compressor fan, pump and compressor) was measured using a power
compresses the working fluid from the superheat state at the analyzer made by WT230, with an accuracy of ±0.02%. The
outlet of evaporator to a high temperature and pressure state air-side cooling capacity can be achieved based on data of
at the compressor outlet (process 1e2). Subsequently, the heat the airflow rate and different sensors. A steady state was
will be rejected to the ambient in the condenser where the maintained for 30 min in psychrometric lab to gather data in
working fluid passes from the superheat state to the saturate each test. During the operation, the thermodynamic prop-
state at the inlet of the receiver (process 3e4). Finally, the erties of the working fluids in cycle were calculated by
saturated or subcooled state liquid refrigerant is expanded REFPROP (Lemmon et al., 2007) using measured tempera-
through thermal expansion valve into the evaporator at a tures and pressures values. All the measurement data was
lower evaporation pressure to cool the data center (process 6- recorded and stored on a computer by using a HP 75000 se-
7-8e1). This cycle is characterized by a high condensing ries data acquisition system. For steady state performance
temperature and a medium overall efficiency when compared tests, the sampling interval was 10 s.
with the liquid refrigerant pump cycle. This is a good opera- The refrigerant enthalpy was determined by the local
tion option when the ambient temperature is high, typically pressure and temperature measurement except the two-
during the summer season. phase mixture state. The air enthalpy was determined by
During the experimental test, a 7 kW fin-tube evaporator the measured dry bulb temperature and wet bulb temperature
was designed to transfer the heat generated by the servers to at the corresponding location.
the refrigerant. The evaporator used is a cross-flow heat The refrigerant-side cooling capacity is calculated as:
exchanger and is composed of four circuits. Each circuit has a
single inlet and outlet. The coil geometry used is summarized Qe;ref ¼ m_ ref ho;ref hi;ref (1)
in Table 1. The condenser used is a cross-flow fin-tube heat where hi,ref is determined by the pressure and temperature
exchanger with two rows and two circuits, and each circuit ahead of the expansion device. ho,ref is determined by the
has a single inlet and outlet. pressure and temperature leaving the evaporator. It should be
Experiments were conducted in a standard psychromet- noted that the refrigerant-side cooling capacity cannot be
ric chambers with two insulated rooms to evaluate the determined when two-phase flow occurs at the exit of an
performance of the designed cycle under a range of oper- evaporator.
ating conditions. The temperature and humidity were The air-side total cooling capacity is calculated as:
independently controlled within each room to the desired
operation conditions. The evaporator air stream was con- Qe;air ¼ m_ air ho;air hi;air (2)
nected to an ASHRAE standard air measurement box to yield
where hi,air and ho,air are determined using the measured inlet
a measurement of the air flow rate and the evaporator air
and outlet air dry bulb and wet bulb temperature of the
flow rate was altered by a variable speed blower inside the
evaporator, respectively.
box. Two pressure transducers are used to measure the
The air-side total sensible cooling capacity is calculated as:
pressure drop across the nozzle and the pressure at the
nozzle inlet, respectively. Three 1000U platinum TEP series Qsen ¼ m_ air ti;air to;air (3)
probes RTDs measure the air temperature near the inlet of
where ti,air and to,air are determined using the measured inlet
the nozzles. All temperature measurements were Type T
and outlet air dry bulb temperatures of the evaporator,
copper-constantan thermocouples, which were calibrated
respectively.
using an ice bath and boiling water to make sure the read-
Energy efficiency ratio is given as.
ings were within 0.3 C of the known values. Low pressure
transducers and high pressure transducers were installed in
the refrigerant flow line to measure the refrigerant pres- I For compressor operation mode :
sures of the interested state points with an accuracy of (4)
Qe;air
±0.25% over the full scale range. The refrigerant tempera- EER ¼
Wcomp þ Wevap þ Wcond
tures and pressures were measured at the locations of the
inlet and outlet of each component throughout the system,
i n t e r n a t i o n a l j o u r n a l o f r e f r i g e r a t i o n 4 9 ( 2 0 1 5 ) 1 1 e1 8 15
Fig. 2 e Effect of ambient air temperature on overall system Fig. 3 e Effect of ambient air temperature on overall system
SHR and sensible capacity with the pump cycle mode. capacity with different cycle modes.
4. Conclusion
compressor when the ambient temperature is high and the EngineeringCongress and Exposition, Anaheim, CA, USA.
other is the pump cycle mode driven by the liquid refrigerant Paper 60669.
pump when the ambient temperature is low. Based on the Kline, S.J., McClintock, F.A., 1953. Describing uncertainties in
single-sample experiments. Mech. Eng. 75, 3e8.
multi-purpose test bench, the performance of the novel
Lemmon, E.W., Huber, M.L., McLinden, M.O., 2007. NIST Standard
hybrid cooling system was experimentally evaluated under Reference Database 23: Reference Fluid Thermaldynamic and
various operation conditions and compared with the tradi- Transport Properties-REFPROP, Version 8.0. National Institute
tional compression cycle systems. The effects of the cycle of Standard and Technology. Standard Reference Data
working condition and shift temperature on the overall cycle Program.
performance were investigated. The annual overall energy Lin, L.C., Ponnappan, R., Yerkes, K., 2006. Actively pumped two-
saving was analyzed using Bin hours method based on the phase loop for spray cooling. J. Thermophy. heat Transf. 20,
107e110.
experimental test data. The results showed that the novel
Liu, Y., 1993. Measuring Retrofit Savings in Commercial Buildings
system has a higher energy efficiency ratio as compared to the with Pre-retrofit Utility Billing Data and Post-retrofit Sub-
traditional compressor system when the ambient tempera- metered Data. M.S. Thesis, Energy Systems Laboratory Report,
ture is lower due to the lower input power. The input power no. ESL-TH-93/12-04. Texas A&M University.
the novel system can be decreased by 50% when compared to Marcinichen, J.B., Thome, J.R., Michel, B., 2010. Cooling of
that of the compressor system. The experimental results also microprocessors with micro-evaporation: a novel two-phase
cooling cycle. Int. J. Refrigeration 33, 1264e1276.
indicated that the proposed novel system can obtain nearly
Marcinichen, J.B., Olivier, J.A., Thom, J.R., 2012. On-chip two-
the same cooling capacity as that of the traditional
phase cooling of datacenters: cooling system and energy
compressor system when the ambient air temperature is recovery evaluation. Appl. Therm. Eng. 41, 36e51.
5 C. The pressure drop of the connected pipe between the Mongia, R., Masahiro, K., DiStefano, E., et al., 2006. Small Scale
evaporator and condenser has a significant impact on to the Refrigeration System for Electronics Cooling within a
total system pressure drop due to the two-phase region under Notebook Computer. ITHERM, San Diego, CA.
the working conditions. The novel system EER is higher than Mudawar, I., Bharathan, D., Kelly, K., et al., 2009. Two-phase spray
cooling of hybrid vehicle electronics. IEEE Trans. Components
that of the compressor cycle when the ambient temperature is
Packag. Technol. 32, 501e512.
below 7 C. The comparative annual energy saving analysis Ohadi, M.M., Dessiatoun, S.V., Choo, K., et al., 2012. A comparison
showed that the lower annual bin temperature is, more en- analysis of air, liquid, and two phases cooling of data centers.
ergy will be saved. The proper shift temperature is about 5 C In: 28th IEEE SEMI-THERM Symposium. IEEE, pp. 58e63.
from the system EER, cooling capacity and annual overall Park, Chanwoo, Vallury, Aparna, Zuo, John, et al., 2007.
energy saving point of view. Overall, this study shows the Electronics thermal management using advanced hibrid two-
proposed novel cycle could provide a way to modify the phase loop technology. In: ASME-JSME Thermal Engineering
Summer Heat Transfer Conference, July 8e12, Vancouver,
traditional air-cooled compressor system and show strong
British Columbia, Canada. HT2007-321962.
energy saving potential for its application in green data Silk, E.A., Golliher, E.L., Selvam, R.P., 2008. Spray cooling heat
centers. transfer: technology overview and assessment of future
challenges for micro-gravity application. Energy Cover.
Manag. 49, 453e468.
references Trutassanawin, S., Groll, E.A., Garimella, S.V., et al., 2006.
Experimental investigation of a miniature-scale refrigeration
system for electronics cooling. IEEE Transact. Components
Brown, R., 2008. Report to Congress on Server and Data Center Packag. Technol. 29 (3), 267e687.
Energy Efficiency Public Law 109-431. U.S. Lawrence Berkeley Wu, D., Marcinichen, J.B., Thome, J.R., 2013. Experimental
National Laboratory. evaluation of a controlled hybrid two-phase multi-
Crepinsek, Michael, Park, Chanwoo, 2012. Experimental analysis microchannel cooling and heat recovery system driven by
of pump-assisted and capillary-driven dual-evaporators two- liquid pump and vapor compressor. Int. J. Refrigeration 36,
phase cooling loop. Appl. Therm. Eng. 38, 133e142. 375e385.
Glassman, B., Kuravi, S., Du, J., et al., 2004. A fluid management Zhou, R., Zhang, T., Catano, J., Tong, T., Chang, J., Prasher, R.,
system for a multiple nozzle array spray cooler. In: 37th AIAA Jensen, M.K., 2010. The steady-state modeling and
Thermo-physics Conference, Portland, Oregon, 2004-2574. optimization of a refrigeration system for high heat flux
Hannemann, R., Marsala, J., Pitasi, M., 2004. Pumped liquid removal. Appl. Therm. Eng. 30 (16), 2347e2356.
multiphase cooling. In: IMECEe Int. Mechanical