Energy savings of hybrid dew-point evaporative cooler and micro-channel separated heat pipe cooling systems for computer data centers

The world has entered the Age of Big Data with large data centers consuming large amounts of energy. 30% e 50% of the energy delivered to a typical data center is consumed by the space cooling system. Dew-point evaporative coolers and heat pipes both utilizing natural cold resources can signi ﬁ cantly reduce these refrigeration costs. This paper presents two hybrid cooling systems combining dew-point evapo-rative coolers with heat pipes for computing and data center cooling systems. The energy-saving potentials of two these hybrid cooling systems were analysed through calculations with comparisons with a traditional vapour compression refrigeration system. The results show that the average annual co-ef ﬁ cients of performance (COP) of the ideal hybrid refrigeration systems are 33 and 34 which leads to annual energy savings of nearly 90% compared with vapour compression refrigeration.


Introduction
Computing and data centers (CDCs) with buildings, services and computing and data processing facilities (e.g. servers, computing telecommunication equipment, air conditioners and power equipment, [1]) have rapidly developed over the past 40 years. In 2010, the total electrical energy used in CDCs was around 1.3% of the world's total energy consumption, with a rate of 1.4% in Europe and 1.5% in China [1e4]. Europe currently has 1148 data centers [5] that consume more than 100 TWh of electricity each year. In China, the CDC capacity reached 28.5 GW in 2013 [3,6], with 549. 6 TWh annual electrical consumption. The CDC capacity continues to expand globally and the CDC electrical energy consumption will continue to grow at an annual rate of 15e20% in the foreseeable future [7].
The Power Usage Effectiveness (PUE), which is defined as the ratio of the total energy used by CDC facilities to the energy delivered to the computing equipment, is about 1.9 in the USA, 2.53 in Europe and 2.5 in China [1e4]. This means that 40%e60% of the electricity delivered to CDCs is used for operating the service facilities (largely for air conditioning), so the CDC operations are very inefficient. Space cooling is a fundamental need of CDCs to remove the tremendous amount of heat dissipated by the IT equipment to keep an adequate space temperature which consumes 30%e40% of energy delivered to the CDCs [1e4]. Thus, CDCs need very efficient cooling systems.
The most common cooling equipment for CDCs is mechanical vapour compression air conditioners, which makes use of high grade energy (i.e., electricity) at relatively low efficiencies (i.e. Coefficients of Performance (COP) of 2e3), leading to an environmentally unfriendly operation.
Since CDCs operate continually every day of the year, natural cooling resources can be used to save energy. Natural cooling resources can be used by direct cooling and indirect cooling systems. Direct cooling systems bring outdoor air directly into the room, which may result in humidity problems and increased pollutants and dust in the CDCs. Indirect cooling systems separate the indoor and outdoor air streams with a heat exchanger such as an air-air heat exchanger or heat pipes, which ensure the humidity and cleanliness requirements of the indoor air.
Heat pipes are very effective heat transfer devices that can efficiently transmit heat over a long distance through a small area without additional power. Heat pipes can be divided into the integral and separated types. Separated heat pipes are much more mature and more widely used than integral heat pipes in CDC cooling systems. However, heat pipe cooling can only be used when the outdoor temperature is below the required temperature and need to be combined with other cooling systems, usually mechanical vapour compression cooling systems, to provide cooling throughout the year. Different kinds of heat pipe systems that have been proposed for CDC cooling systems are summarized in Table 1.
Although heat pipe-vapour compression compound cooling systems save much energy compared to simple mechanical vapour compression cooling systems, the electricity energy consumption is still large when the outdoor air temperature is high, and needs to be further reduced by new cooling methods.
Evaporative cooling can conserve energy and protect the environment by using water evaporation to efficiently absorb the heat [18]. The existing evaporative cooling systems can be classified into direct evaporative cooling (DEC) and indirect evaporative cooling (IEC) systems. Indirect evaporative cooling system, where the product air in dry channel and working air and water film in wet channel are separated by heat exchange plate, can lower the product air temperature without adding moisture to product air, so they are more suitable for air conditioning in more climates than DEC systems and, thus, are more widely used [19e21]. Many evaporative cooling systems have been used for CDC cooling like direct evaporative cooling system [22], indirect evaporative cooling with direct expansion refrigeration system [23], and indirect evaporative cooling with dew-point evaporative cooling system [24]. However, evaporative cooling systems are not widely used because of their relative low energy efficiencies, large volumes and restrictions on the supply air temperature [25]. A counter-flow corrugated plate dew-point evaporative cooler is described in this paper that gives higher cooling efficiencies (wet-bulb efficiency) and energy efficiency (COP) than ordinary indirect evaporative cooling by improving the design of the indirect evaporative cooling system. Experimental results show that the cooling efficiency can reach 125% and the COP can reach 43 for a 10 kW dew-point evaporative cooler. Thus, dew-point evaporative cooling has great potential for saving energy in CDC cooling. However, the dew-point evaporative cooler, as a standalone system with an air exhaust tube, can only supply air centrally, e.g. the upper air supply or the downside air supply, while the evaporators with the separated heat pipes can be distributed to each IT system to further improve the heat transfer efficiency.
The heat pipe and dew-point evaporative cooler are both energy efficient and environmentally friendly but the heat pipe can only be used when the outdoor temperature is below a certain temperature while the dew-point evaporative cooler can give high cooling efficiencies even with high outdoor temperatures. Therefore, a heat pipe-dew-point evaporative cooler hybrid cooling system is proposed and analysed here to satisfy the CDC cooling requirements and simultaneously save enormous amounts of energy. This paper describes two hybrid cooling systems combining micro-channel separated heat pipes with dew-point evaporative coolers with the energy saving potentials of the two hybrid cooling systems analysed through calculations.

CDC indoor air requirements
According to the ChineseIStandard Specifications for Design of Electronic Information System RoomsJ(GB 50174-2008) [26], data centers are classified into three levels based on their functions, significance and reliability. Generally, large data centers are classified into level A (extremely important) and B (important). The environmental specifications for level A and B data centers are shown in Table 2. The following designs are based on the specifications listed in Table 2 [26].
The local air supplies are normally separated into cold and hot aisles in CDC cooling systems. Considering the improvement of the performance of IT equipment for the state-of-the art data center, the cold aisle temperature was assumed to be 23 ± 1 C, and the hot aisle temperature was set to be 35 C to save energy.

CDC operating assumptions
The CDC IT equipment is assumed to dissipate 5 kW per piece and 100 kW in total. Thus, the CDC is assumed to have: 100 kW=5 kW ¼ 20 pieces (1) According to the ChineseIStandard Specifications for Design of Electronic Information System RoomsJ(GB 50174-2008), the floor area of CDC room before the equipment is selected can be estimated as: where K is the floor area for each piece of equipment, usually 3.5e5.5 m 2 and N is the number of computer systems in the room. The floor area for each computer system is set to 5 m 2 in this study, so the total area of the CDC is 100 m 2 .
The thermal load coefficient of outside environment is assumed to be 0.15 kW/m 2 , so the total thermal load from outside environment is 15 kW and the total thermal load that needs to be removed by cooling system of the CDC is 115 kW, which is the base for Table 1 Recent heat pipe research for CDC cooling systems.

Researcher
Heat Pipe System Year Weber and Wyatt [8] Integrated heat pipe system with a pressurized air duct 2011 Xia et al. [9] Separated heat pipe system with forced convection 2008 Xia et al. [10] Separated heat pipe system with natural convection 2008 Suenaga and Ichimura [11] Heat pipe-vapour compression one-piece air conditioner 1986 Wang et al. [12] Heat pipe-vapour compression compound air conditioner with additional heat exchanger 2013 Han et al. [13] Heat pipe-vapour compression compound air conditioner with three-way valve 2013 Yan et al. [14], Ma et al. [15] Heat pipe-vapour compression compound air conditioner driven by a liquid pump 2015 Zhang et al. [16] Mechanical refrigeration/loop heat pipe one-piece air conditioner based on a three-medium heat exchanger 2015 Lee et al. [17] Heat pipe-vapour compression compound air conditioner with four solenoid valves 2006 Table 2 Environmental specification for data centers.
-A and B levels Temperature/ C 2 3 ± 1 Relative Humidity/% 40e50 Temperature variation rate/( C/h) <5, moisture condensation not allowed designing the hybrid cooling system.

Hybrid cooling system design
Two different hybrid cooling systems with separated heat pipes and dew-point evaporative coolers were designed to satisfy the year round CDC cooling requirements with the designs shown in Figs. 1 and 2. The evaporators (2) of the separated heat pipes are distributed directly to the backs of the IT equipment (10) and are equipped with variable frequency fans to adjust the air volume according to the system operating conditions. A liquid pump (5) pumps liquid around the bypass in the liquid line of the heat pipe system when the outdoor temperature is very high; otherwise, the heat pipe works in natural circulation mode. Indoor air flows through the IT equipment (10) where it is heated by the IT equipment and then flows through the evaporators (2) on the back of the IT equipment where it is again cooled to the indoor air temperature. The evaporator layout is shown in Fig. 3. The dew-point evaporative cooler designs differ in the two systems with the specific operating modes for the two systems are listed in Tables 3 and 4.
System A: The heat pipe and the dew-point evaporative cooler can run independently or simultaneously. When the outdoor air temperature is below 12 C, the system uses only the heat pipe.
When the outdoor air temperature is between 12 and 18 C, the heat pipe alone can no longer satisfy the cooling requirements, so the dew-point evaporative cooler is also run. The supply air from the dew-point evaporative cooler is delivered to the room through the bottom air duct (9). When the outdoor air temperature is above 18 C, the heat pipe cannot work and only the dew-point evaporative cooler is used.
System B: When the outdoor air temperature is below 12 C, only the heat pipe is used as in system A. When the outdoor air temperature is between 12 and 25 C, the heat pipe alone can no longer satisfy the cooling requirements, so supply air from the dewpoint evaporative cooler is used as the cooling air for the heat pipe condenser in the cascade mode. When the outdoor air temperature is between 25 and 26 C, the cascade mode can no longer satisfy the cooling requirements, so part of the supply air from the dew-point evaporative cooler is delivered to the room through the bottom air duct for cooling. When the outdoor air temperature is over 27 C, the heat pipe cannot work and only the dew-point evaporative cooler is used.

Design of the micro-channel separated heat pipe
The flow diagram for the separated heat pipe is shown in Fig. 4 with a liquid pump in the liquid tube to supplement the natural    Compound mode >18 Dew-point evaporative cooler mode Table 4 System B operating modes.
Outdoor air temperature/ C Operating mode 12 Heat pipe mode 12e25 Cascade mode 25e26 Compound and cascade mode !27 Dew-point evaporative cooler mode convection flow when needed. When the temperature difference between the indoor and outdoor air is too large, the natural circulation mode will be changed to the circulation mode driven by the liquid pump to improve the cycle performance. The heat pipe was designed based on "ideal cycle" assumptions as follows: (1) Only two-phase flow exists in the evaporator and condenser.
(2) The evaporation temperature is equal to the condensation temperature.

Micro-channel heat exchanger size
A micro-channel heat exchanger can increase the heat transfer coefficient and the heat transfer rate compared to conventional size heat exchangers. The sizes of the flat tube and the louvered fins for the evaporator and condenser in the heat pipe are shown in Tables 5 and 6 with a schematic of the micro-channel flat tube section shown in Fig. 5.
The air side heat exchange area for a single-row, micro-channel heat exchanger per unit length is: where N is the number of flat tubes. The refrigerant side heat exchange area for the single-row, micro-channel heat exchanger per unit length is: The contact area between the air and the fin in the single-row, micro-channel heat exchanger per unit length is:

Condenser design
The volumetric airflow rate and the combination with the dewpoint evaporative coolers will require five condensers with the heat load of each condenser (Q c ) assumed to be 23 kW R134a is used as the refrigerant and the assumed design parameters are listed in Table 7. The air side heat transfer coefficient, h a,c , is given by the correlation proposed by Kim and Bullard [27].
where the Colburn j-factor is calculated as: The Fanning friction factor is given by, where Re Lp is the air-side Reynolds number based on the louver pitch and is calculated as: The air velocity between the fins, u a , can be calculated based on the given face velocity of the air, u, and the heat exchanger design parameters: The surface effectiveness, h a;c , and the fin efficiency, h f , on the air side are given by: [28]. where, The heat transfer calculation on the refrigerant side assumes that the refrigerant is two-phase in the condenser, but actually the refrigerant is superheated at the condenser inlet and is supercooled at the condenser outlet. Thus, the calculated heat transfer coefficient will be lower than the actual coefficient, but the difference is small and can be neglected. The refrigerant side condensation heat transfer coefficient, h r,c , is calculated using the correlation proposed by Akers [29].
where Re eq is the equivalent all liquid Reynolds number for the refrigerant calculated as: where Geq is the equivalent all liquid mass flow velocity of the refrigerant, D h is the flat tube equivalent diameter, Gc is the mass flow velocity of the refrigerant, q r,c is the mass flux of refrigerant calculated by, where Q c is the heat load of each condenser and h fg is latent heat of the refrigerant.
The surface effectiveness, h r;c , and the fin efficiency, h b , on the refrigerant side are calculated as: [30].
where A bo is the contact area between the refrigerant and the slices of a single flat tube, The total heat transfer coefficient for the condenser can then be calculated as: Then, the total heat transfer area and the heat exchanger sizes can be calculated as: where q m;c is the heat transfer temperature difference given by: The final results for the condenser are shown in Table 8.

Evaporator design
The evaporators are located on the back of the IT equipment as shown in Fig. 3. The design heat load of each evaporator is 5.75 kW. The temperature of air distributed to the IT equipment is 23 ± 1 C and increases to 35 C after flowing through the IT equipment. Then, the air flows through the heat pipe evaporator and is cooled to 22 C again. The evaporation temperature is set to 20 C to avoid moisture condensation. The design parameters are listed in Table 9.
The volumetric airflow rate through each evaporator is: The air-side heat transfer coefficient for the evaporator is calculated using Eqs. (6)e (14). The results for the air-side heat transfer are shown in Table 10.
The boiling heat transfer coefficient in the micro-channels is calculated using the correlation proposed by Sun and Mishima [30].
where We l is the Weber number for the liquid refrigerant defined as: Bo is the Boiling number defined as: Re lo is the equivalent all liquid Reynolds number defined as: The total evaporator heat transfer coefficient is then defined as: The total outside heat transfer area is: The heat flux inside the tube, q i , can then be calculated and checked against the required heat transfer rate: The final results for the evaporator are shown in Table 11.

Pressure drop check
The heat pipe condenser and evaporator designs are based on the assumption that the refrigerant flows with negligible pressure loss so the condensation temperature equals the evaporation temperature. The pressure drops should then be checked to verify that the design is reasonable.
The heat pipe evaporator was assumed to be 2.8 m below the condenser according to the CDC construction standard. The heat pipe heat transfer rate with natural circulation is influenced by the tube diameter of the rising vapour tube with a large diameter needed to work efficiently [31]. Thus, the diameters of the rising vapour tube and the falling liquid tube were both set to 25 mm. The rising vapour tube is 3.1 m high and the falling liquid tube is 3 m high.
The equilibrium equation for the pressure drops for natural circulation is:  Table 9 Evaporator design parameters.
Evaporation temperature t e ( C) Inlet air dry bulb temperature t 1 e ( C) Outlet air dry bulb temperature t 2 e ( C) Indoor air relative humidity (%) 20 35 22 45 DP g À DP fr À DP ac À DP local c þ DP g À DP fr À DP local fa ¼ DP g þ DP fr þ DP ac þ DP local e þ DP g þ DP fr þ DP local ri (37) The pressure drops in the vapour tube and the liquid tube are calculated using the single-phase pressure drop equation: À dp dz ¼ rg sin q þ dp fr þ dp local The three terms on the right side represent the gravitational pressure drop, the friction pressure drop and the minor losses. The acceleration pressure drop is ignored because the density changes little for single-phase flow. The minor pressure drop includes the pressure drop for the refrigerant flow through the two elbows in the tube line. The liquid tube pressure drop includes the pressure drops in the liquid part and the vapour part. The height of liquid part can be calculated using equation (37). The pressure drops are then listed in Table 12.
The two-phase flow pressure drops in the evaporator and condenser are calculated as: The three terms on the right represent the gravitational pressure drop, the acceleration pressure drop, the friction pressure drop and the minor losses. The void fraction, a, is calculated using the Thom model [32]: The friction pressure drop is calculated using the correlation proposed by Fridel [33]. The two-phase multiplier is given by: Fr 0:045 We 0:035 (41) The parameters E, F, and X are defined as: The Froude and Weber numbers are defined as: The two-phase mixture density is given by: The terms with subscripts l0 and g0 correspond to the frictional pressure gradients when all the mixture is either liquid or gas. f g0 and f l0 are calculated using the Churchill correlation: [34].
The calculated pressure drops in the evaporator and condenser are listed in Table 13. The results show that the assumption that the heat pipe condensing and evaporation temperatures are the same is reasonable.
The condensation temperature is then calculated after taking  into account the pressure drops with the change in the condensation temperature shown in Table 14.

Dew-point evaporative cooler design
The 10 kW dew-point cooler was designed based on numerical simulations of a dew-point evaporative cooler. The heat exchanger is made of corrugated plates to increase the heat and mass transfer surface area and the wet surface is covered with a porous fabric to improve the water diffusion and increase the contact area between the wet air and the water. The design parameters are listed in Table 15. The dew-point evaporative cooler includes the heat exchanger, supply fan, exhaust fan, circulation water pump, water distributor and water tank as shown in Fig. 6.
Tests have shown that the supply air temperature is below 18 C, the cooling efficiency (Wet-bulb efficiencyε) can reach 125% and the cooler COP can reach 31 for a 10 kW dew-point evaporative cooler for outdoor with a dry-bulb temperature of 38.2 C and a wet-bulb temperature of 22.1 C. The detailed experimental results which has not been published are shown in Table 16. The performance of dew-point evaporative cooler is influenced by the outdoor air humidity, so the dehumidification is needed in highhumidity weather to guarantee that the temperature and humidity of the supply air delivered to the room are satisfactory. According to the experimental and simulation results, the changes of dry-bulb temperature of supply air are minor with the verification of drybulb temperature of inlet air if the dew-point temperature of inlet air is constant for the dew-point evaporative cooler. The dewpoint temperature of air is influenced by humidity of air only, so the supply air temperature and cooling capacity of the dew-point evaporative cooler can be controlled by controlling the humidity of inlet air.
This paper focuses the energy savings of the two hybrid cooling systems, so the cooling by the single dew-point evaporative cooler is assumed to be the same as in experimental results for the two systems to simplify the calculations with the energy consumed by the dew-point evaporative coolers being a function of the number of dew-point evaporative coolers required by the system. The detailed parameters for the dew-point evaporative cooler used in the systems are shown in Table 17.

Energy-saving analyses of the hybrid cooling systems
The dry-bulb temperature distribution in Beijing is shown in Fig. 7 [35]. There are on average 4091.4 h per year with dry-bulb temperatures below 12 C which is 46.7% of the time. There are then about 3539.4 h per year with dry-bulb temperatures over 18 C which is 40.4% of the time. These numbers were then used to calculate the annual operating hours for each hybrid cooling system operating mode listed in Tables 3 and 4 with the results shown in Fig. 8.
The energy consumption in the heat pipe and the dew-point evaporative cooler systems is mainly the energy consumed by the fans. Thus, the system energy consumption is calculated as: where DP is the air pressure drop, V is the volumetric air flowrate, h 1 is the fan efficiency, h 2 is the mechanical efficiency, and M C is the motor capacity storage coefficient. The air pressure drops across the heat pipe evaporator and condenser are calculated by: where the fanning friction factor, f, is calculated using Eq. (8) [29].
A a0 is the air side heat exchanger area, calculated using Eq. (3) A c0 is the minimum cross-sectional surface area u c is the velocity at the minimum cross-sectional area The energy consumption of the dew-point cooler is calculated based on the parameters in Table 17 and the number of dew-point evaporative coolers used in the hybrid cooling system. As shown in Table 17, the power consumed per dew-point evaporative cooler is 0.32 kW.
The total energy consumption of the cooling systems is listed in Tables 18 and 19. The system COP of the heat pipe working alone can reach 42 with the annual average COP for hybrid system A being 33 and the for system B being 34. The total energy consumption rates for the two hybrid systems are compared with that of a conventional vapour compression refrigeration system in Table 20. The results show that the hybrid cooling systems described in this paper are much more efficient than conventional vapour compression refrigeration systems with and the energysaving ratios of nearly 90%. Thus, the hybrid cooling systems can significantly reduce the CDC cooling cost. System B is more efficient than system A, but system B is more complicated than system A which will increase the operating and investment costs of system B. Thus, further life-cycle costs are needed to determine which CDC cooling system should be used.
In addition, a dehumidification system is needed when the dewpoint evaporative coolers work, so the actual energy consumption will be larger than the calculated results.

Conclusions
This paper describes data center cooling systems that effectively use natural cold sources to reduce the energy consumption in CDCs. Two hybrid cooling systems are given which combine dew-point evaporative coolers and micro-channel separated heat pipes which both provide significant energy savings. The micro-channel separated heat pipe is designed in detail and the dew-point evaporative cooler is chosen based on previous studies of dew-point evaporative coolers. Then, the energy consumption rates of both systems are calculated based on Beijing weather conditions. The conclusions are: (1) Two kinds of hybrid cooling systems are described that both combine dew-point evaporative coolers and micro-channel separated heat pipes along with the system operating modes that vary with the outdoor air temperature. The results show that the annual energy consumption is 32016 kW h and the average COP is 33 for system A. The annual energy consumption is 31470 kW h and the average COP is 34 for system B. But the actual energy consumption is larger than calculation results considering the energy consumption of dehumidification system. (5) Both hybrid cooling systems will use nearly 90% less electricity than a conventional vapour compression system. Thus, the hybrid cooling systems combining dew-point   All mixture assumed to be vapour i internal surface in inlet l liquid phase l0 All mixture assumed to be liquid o outside surface r refrigerant ri rising tp two phase