Energy conservation effects of a multi-stage outdoor air enabled cooling system in a data center

Energy and Buildings 138 (2017) 257–270

Contents lists available at ScienceDirect

Energy and Buildings journal homepage: www.elsevier.com/locate/enbuild

Energy conservation effects of a multi-stage outdoor air enabled cooling system in a data center Je-Yong Kim a , Hyun-Jae Chang b,∗ , Yong-Ho Jung c , Kyu-Man Cho d , Godfried Augenbroe e a

Graduate School, Hongik University, Jochiwon, Sejong, Korea School of Architectural Engineering, Hongik University, Jochiwon, Sejong, Korea Dept. of Building Equipment and Fire Protection System, Chungwoon University, Incheon, Korea d Dept. of Architectural Engineering, Chosun University, Gwangju, Korea e College of Architecture, Georgia Institute of Technology, Atlanta, GA, USA b c

a r t i c l e

i n f o

Article history: Received 26 May 2016 Received in revised form 5 December 2016 Accepted 18 December 2016 Available online 21 December 2016 Keywords: Data center Economizer cycles MOA cooling system Energy simulation

a b s t r a c t Energy consumption rates in a data center have increased year by year, and by this, the introduction of energy conserving cooling systems is rapidly emerging all over the world. In this study, a multi-stage outdoor air enabled (MOA) cooling system is proposed as the safer and more energy conserving system for a data center. An MOA cooling system is composed of the combination of water-side economizer, airside economizer and mechanical cooling. Effects of an MOA cooling system are investigated by energy simulation from various viewpoints. The results show that the MOA cooling system can save cooling energy consumption rates by about 21%, and if combined with a special rack enclosure, by about 33%. Introducing outdoor air flow into the MOA cooling system decreases energy use to about half of that of the air-side economizer, and if combined with the rack enclosure, by about 9%. Thus, the concerns about particulate and gaseous contamination can be reduced. Continuous operating time of the humidifier drops to about 3 days, and if combined with the rack enclosure, to ‘0’ day. These results make it possible to remove the humidifier as part of the MOA cooling system. © 2016 Elsevier B.V. All rights reserved.

1. Introduction With the rapid development of IT (information technology) industries over recent decades, the needs for data centers has grown steadily. Server rooms in a large data center are composed of many server racks which house server computers, storage and other systems. Heat generation rates per rack have increased year by year through the improvement of server computer performance, and it is anticipated to become even larger with the appearance of the blade server. Based on predictions, the power consumption rates of data centers constitutes over 0.5% of the total power use around the globe [1]. The data center market is growing very fast. Michael Rutberg et al. mentioned by citing a 2011 report from Microsoft that annual market size for data center construction was about $15 billion for the U.S. and $50 billion globally, and will be growing to $80 billion globally by 2020 [2,3]. Cooling energy need in a data center accounts for about 30% of the energy consumption of an average data center [4,5]. There

∗ Corresponding author. E-mail address: [email protected] (H.-J. Chang). http://dx.doi.org/10.1016/j.enbuild.2016.12.057 0378-7788/© 2016 Elsevier B.V. All rights reserved.

may be several ways to conserve cooling energy. Economizer cycles using free cooling can be an effective energy saving method. Efforts to reduce cooling energy consumption by economizer cycles have been conducted by many researchers. Steven Pelley et al. developed an analytic framework for modeling total data center power and abstract models that replace key simulation steps with simple parametric models [6]. Jinkyun Cho et al. compared the energy performance of a water-side economizer and air-side economizer in temperate or subtropical regions, and estimated their improvement of energy performance at 16.6% and at 42.4% in each [7]. Hao Tian et al. investigated the water-side free cooling potential in a combined system of multi-stage heat pipe and water loop with serially connected multi cold sources. They found that energy efficient cooling could be expected, and confirmed that the annual cooling cost reduction is approximately 46% through a comparative measurement [8]. Christy Sujatha.D et al. estimated energy consumption in a data center using a conventional cooling system, air-side economizer and water-side economizer for three different zones whose outside air is relatively cool for most of the year, and found that water-side economizers were consistently outperforming an air-side economizer [9]. Aayush Agrawal et al. examined energy savings potential of water-side economizer and indirect

258

J.-Y. Kim et al. / Energy and Buildings 138 (2017) 257–270

evaporative cooler for the 17 worldwide climate zones defined by ASHRAE 90.1, and found that more than 30% energy savings could be achieved by indirect evaporative cooler in climate zones 3–7, and more than 25% by water-side economizer in climate zones 4 and 5 [10]. Bruce A. Hellmer analyzed the electrical and water consumption for cooling systems of telco/data centers located in various cities throughout the United States, and found that water-side economizers are less effective than air-side economizers at improving energy efficiency and reducing refrigeration dependency. They also found that air-side economizers with steam humidifiers are an energy wasteful design because more power was consumed by the steam humidifiers than is saved by the air-side economizers [11]. Jayantha Siriwardana et al. investigated cooling potential of airside economizers in 20 cities of Australia, and demonstrated sizable potential for significant savings on cooling costs [12]. Kuei-peng Lee et al. examined the potential energy savings by air-side economizer with differential enthalpy control in 17 world climate zones. Their results show that energy consumption of fan and the humidification/dehumidification system substantially reduced the benefit of free cooling, and therefore, the air-side economizer could not be an optimal or suitable solution in some climate zones [13]. SangWoo Ham et al. investigated an energy optimization process for the air-side economizer in a modular data center with respect to various parameters, and suggested the optimum supply air temperature range of 18–23 ◦ C [14]. They also studied the applicability of various air-side economizers in modular data centers, and found that total cooling energy savings of 47.5% to 67.2% were possible [15], and in an examination of water-side economizer and cold aisle containment in a modular datacenter, they found that the energy savings of a water-side economizer combined with cold aisle containment are much greater than without containment [16]. Milnes P. David et al. tested a chiller-less data center experimentally where server-level cooling is achieved through a combination of warm water cooling hardware and re-circulated air, and characterized an average cooling energy use of 3.5% of the total IT energy use [17]. Hiroshi Endo et al. examined the use of fresh-air cooling to reduce energy consumption in container data centers. They confirmed that direct fresh-air cooling combined with evaporative cooling and waste heat from IT equipment is available even when the outside air conditions were out of acceptable server setting ranges [18]. Baptiste Durand Estebe et al. investigated the effects of cooling plant energy conservation by a strategy of regulating supply air temperature in the case of an Integrated Water Side Economizer system, and showed that 17% of electric energy used by the cooling plant could be saved [19,20]. Eduard Oró et al. investigated an energy saving potential of the integration of direct air free cooling strategy and thermal energy storage system in data centers located at different European locations, and showed that the operational cooling cost could be drastically reduced when thermal energy storage was used in combination with an off-peak electricity tariff [21]. They also summarized renewable energy integration into data centers and its characterization using numerical models [22]. Ali Habibi Khalaj et al. analyzed the effectiveness of energy and exergy, and the environmental and economic impact of the cooling system with various economizers for 23 locations across Australia. They found that the energy saving potential increased for locations further south in Australia [23]. Jinkyun Cho et al. investigated each technical component’s influence on energy-optimized data centers, and derived how the prioritization of system selection should be made [24]; they proposed an analytical approach methods for the energy efficiency optimization of high density data center for four different climate regions [25]. An air-side economizer is a very efficient system for the cooling energy conservation in a data center. However, there are concerns of particulate contamination and gaseous contamination. Water-

side economizers on the other hand is a safe solution, but its available operation hours are very limited. In the researches so far, an air-side economizer or water-side economizer has been investigated in terms of their energy conservation effect individually or in combination with other mechanical systems. Authors recognize that the merits and demerits of air-side economizers and water-side economizers are contrary to each other. In this study, by integrating the merits of air-side economizers and water-side economizers, a multi-stage outdoor air enabled (MOA) cooling strategy was established, and its cooling energy conservation effects are investigated. The MOA cooling strategy is based on the combination of air-side economizers, water-side economizers and mechanical cooling, and by virtue of the combination, more efficient and less risky cooling for data centers is expected. 2. Economizer cycles Economizer cycles are free cooling systems utilizing outdoor air under certain conditions to allow chillers and/or other mechanical cooling systems to be shut off or operated at reduced capacity [26]. 2.1. Air-side economizers An air-side economizer is an energy conserving system for cooling by direct introduction of outdoor air under proper conditions. Fig. 1 shows the schematic of air-side economizers. In an air-side economizer, outdoor air is introduced to a computer room by computer room air handler (CRAH) units when the enthalpy of outdoor air meets the requirement for cooling. Air-side economizers can be operated fully or partially, and it needs an additional fan for the exhaust air. Comparing with water-side economizers, in general, air-side economizers are more energy efficient, and their available operation is longer. There have been four concerns with the use of air-side economizers in data centers [11] which can be summarized as: - Increased particulate contamination and/or increased maintenance cost on filters. - Increased gaseous contamination. - Loss of humidity control during economizer operation and loss of a vapor seal during non-economizer operation. - Temporary loss of temperature control during economizer switchover with non-integral economizers. A study by LBNL showed that particulate concentrations in data centers could be kept to those of current data centers by applying MERV 11 filters, and that the humidity was within ASHRAE’s recommended range in most data centers [27]. 2.2. Water-side economizers Water-side economizers are operated by the aid of cooling towers. When outdoor wet-bulb temperature meets the facility’s cooling requirements, a computer room can be cooled only by condenser water. There are two types of water-side economizers, i.e. direct and indirect. In case of direct economizer, the condenser water flows into the chilled water circuit. Even though the direct economizer is a more energy efficient system, there may be tube fouling concerns [26]. Indirect economizer needs heat exchanger between condenser water and chilled water, and, therefore, its efficiency is lower than the direct economizer. Fig. 2 shows the schematic of the indirect water-side economizer. Water-side economizers, compared with air-side economizers, are the safer system with no concerns mentioned in the previous section. But their available use is shorter.

J.-Y. Kim et al. / Energy and Buildings 138 (2017) 257–270

259

Fig. 1. Schematic of the air-side economizers.

Fig. 2. Schematic of the water-side economizer (indirect type).

2.3. Multi-stage outdoor air enabled cooling system Air-side economizers and water-side economizers have the merits and demerits which are contrary to each other. If their merits are combined together, the safer and the more efficient economizer cycles may be possible. In this study, by integrating air-side economizers and water-side economizers, a multi-stage outdoor

air enabled (MOA) cooling strategy is established. Fig. 3 shows the schematic of MOA cooling system which is composed of both economizers. Operating conditions of MOA cooling system are as shown in Fig. 4 in the psychometric chart. Dotted air conditions in Fig. 4 show the annual outdoor air conditions in Seoul, Korea. In the MOA cooling system, basically, the water-side economizer cycle is the priority operating mode because it is safer than the use

260

J.-Y. Kim et al. / Energy and Buildings 138 (2017) 257–270

Fig. 3. Schematic of the Multi-stage outdoor air enabled (MOA) cooling system.

of air-side economizers. Zone “W” is the region that the wet-bulb temperature of outdoor air is low enough, and cooling loads can be removed by the water-side economizer alone. Zone “WA” shows the region that cooling capacity by only the water-side economizer is not sufficient, and needs the aid of air-side economizer. In this zone, heat is eliminated by the water-side economizer first, and the remainder is rejected by the air-side economizer. Introduced outdoor air volume by the air-side economizer is controlled by an outdoor air (OA) damper to fit the supply air temperature to the set point. Zone “WA” is introduced as a new addition in this study for reducing outdoor air volume. Zone “A1” is the region that the enthalpy of outdoor air is low enough to operate the airside economizer, but the water-side economizers cannot operate because the wet-bulb temperature of outdoor air is higher than required. The outdoor air temperature is limited to be below the supply air temperature, and dew-point temperature below 15 ◦ C. The cooling load is removed only by the air-side economizer, and the introduced outdoor air volume is controlled to fit the supply air temperature to the set point. Zone “A2” shows the region that the enthalpy and temperature of outdoor air are lower than those of return air, and dew-point temperature is lower than 15 ◦ C. In this zone, 100% outdoor air is brought in, and the mechanical cooling system is operated to lower its temperature. Zone “M” shows the region that economizer cycles cannot operate because the enthalpy and temperature of outdoor air are higher than those of return air, and mechanical cooling system is operated fully.

3. Rack enclosure In a computer room of data centers, there may be an air mix between cold air supplied by the computer room air conditioner (CRAC) units and hot air exhausted from the servers. J. Cho and et al. investigated air recirculation that occurs at the top of the racks due to an insufficient supply flow rate [28]. ASHRAE pointed out the

following potential problems when air is mixed by recirculation and bypass [29],

• The IT equipment consumes more fan power to compensate for the higher inlet temperature. • The air temperature supplied to the IT equipment is increased. This may require that CRAC unit set points be lowered to maintain a suitable inlet temperature, or to avoid hot spots. • The air temperature returned to the CRAC unit is lowered. This lowers the overall cooling capacity of the CRAC unit and may require additional airflow (or additional CRAC units) to compensate.

Rack enclosure is a system for preventing unwanted air mix in a computer room as an integral part of the airflow system of data centers [30]. By utilizing rack enclosures, the cold air supplied by CRAC or computer room air handler (CRAH) unit can be separated perfectly from the hot air exhausted from the server computers. By this, it is possible to increase rack supply air temperature and achieve the same cooling effect, and this makes the followings possible [1];

• Reduced cooling demand at the central building cooling plant • Increased number of hours per year that the data center can be cooled by economizer cycles; either air side or water side. • Reduced heat gains from thermal conductivity through walls due to smaller temperature difference between inside and outside

Additionally, it is able to improve the COP of the chiller. As shown in Fig. 5, if the leaving chilled water temperature becomes higher, the chiller efficiency can increase dramatically [31].

J.-Y. Kim et al. / Energy and Buildings 138 (2017) 257–270

261

Fig. 4. Operation modes of multi-stage outdoor air enabled (MOA) cooling system.

10.0

Table 1 Configurations of the test bed.

Chiller efficiency (COP)

9.5 9.0 8.5 8.0 7.5 7.0

Location No. of racks

Hongik university, Sejong city, South Korea 10 racks

Rack arrangement Heat generation rates Cooling system

Cold aisle and hot aisle Max. 5 kW/rack Mechanical system + water-side economizer + air-side economizer Underfloor plenum supply/Ceiling plenum return

Air inlet/outlet

6.5 6.0 5.5 5.0

7.0

9.0

11.0

13.0

15.0

17.0

Leaving water temperature (deg C) Fig. 5. Chiller efficiency as a function of leaving chilled water temperature.

4. Methodologies 4.1. Verification of energy simulation methods In this study, for the purposes of estimating energy conservation effects of an MOA cooling system, an experimental study and simulation study were conducted in series. A testbed for a data center, as shown in Fig. 6, was constructed for the experimental study. Table 1 shows the configurations of the testbed, and Table 2 shows the specifications of the cooling system. In a server room, 10 server racks whose maximum heat generation rates are

5 kW/rack are arranged according to the cold aisle/hot aisle layout [32]. The operable modes for cooling are mechanical cooling, water-side economizer cycles, air-side economizer cycles and MOA cooling that is an integrated cooling strategy. Airflow volumes of supply air, return air, exhaust air and outdoor air were monitored in real time by a flow measuring system (FMS). Temperature and humidity of supply air, return air, exhaust air, outdoor air and air at rack inlet/outlet are monitored in real time as well. Experimental studies for evaluating energy consumption rates were conducted in each operating mode for a short time, typically 24 h. The results were compared with the results of energy simulation to verify the simulation methods. An energy simulation was conducted by utilizing TRNSYS 17. Based on the experimental results of cooling energy consumption rates in the testbed, the simulation approach was examined. Fig. 7 shows the comparison of real time cooling energy consumption rates between experiment and simulation for the testbed. Simulation results show good consistency with the experimental results.

262

J.-Y. Kim et al. / Energy and Buildings 138 (2017) 257–270

Fig. 6. Test bed for a data center.

Table 2 Specifications of cooling system of the test bed. Equipment

Quantity

Capacity

Control mode

Chiller Closed circuit cooling tower Chilled water pump Condenser water pump CRAH unit Exhaust fan Humidifier

1 1 1 1 1 1 1

45,000 kcal/h (Scroll type, COP: 3.4 at LW Temp.†1 7 ◦ C) 45,000 kcal/h (LW Temp.†1 : 32 ◦ C) 165 LPM 165 LPM 45,000 kcal/h (Air volume: 9,600 CMH) 9,600 CMH 15 kg/h

ON/OFF control ON/OFF control Full operation/ OFF in air-side mode ON/OFF control CAV CAV (damper control) ON/OFF control

LW Temp.†1 : Leaving water temperature.

Table 4 Heat generation schedule of server computers.

Table 3 Configurations of the target data center. Floor area No. of server computers Heat gains from Server computers Lighting People

2,250m2 × 9 story 720EA/floor × 9 story 2,880 kW/floor × 9 story 49.5 kW/floor × 9 story Sensible heat: 60W/person × 20persons/floor × 9 story Latent heat: 40W/person × 20persons/floor × 9 story

It is thought that this good consistency results from the fact that indoor heat generation rates from the server racks (50 kW) are constant and much larger than other cooling loads. Fig. 8 shows the comparison of the total cooling energy consumption during the test period between experimental studies and simulation studies. In each operating mode, the maximum error was 6.3% in case of airside economizer (A1 + A2). And, in cases of changing outlet water temperature and supply air temperature, the maximum error was 5.3% in case that outlet water temperature is 7 ◦ C, and supply air temperature is 13 ◦ C. Based on these results, the energy simulation approach was confirmed as effective. The same simulation methods were consequently scaled up to a large scale data center. 4.2. Energy simulation for a large scale data center The target data center for a large scale simulation is located in Seoul, Korea. Table 3 shows the configuration of the target data center. It is a 9 story building whose floor area of a computer room is 2,250 m2 . There are 720 server racks per floor, and heat generation rates are 4 kW/rack. The heat generation schedule of server computers is as shown in Table 4. Specifications of the cooling system are as shown in Table 5. Each equipment is controlled under the control mode as shown in Table 5.

Time

Weekday (%)

Weekend (%)

00:00–09:00 09:00–21:00 21:00–24:00

90 100 90

80 90 80

Fig. 9 shows the weather conditions in Seoul from TMY2 where the target data center is located. Seoul is in the temperate climate region, and it shows high temperatures and humidity in summer season, and low temperatures and humidity in winter season [33]. 5. Results and discussion 5.1. Effects of MOA cooling system Energy simulations were conducted for the cases of the MOA cooling system, mechanical cooling system and the economizer cycles for comparison purposes. Table 6 shows the specific outdoor air conditions for each operating mode of the MOA cooling system shown in Fig. 4. These conditions correspond to the outdoor air conditions for the economizer cycles. Table 7 shows the simulation results in each operating mode. Denotation of “ M” means that it needs the help of the mechanical cooling. Mode “M” is the traditional cooling mode, i.e. when cooling loads are removed only by the mechanical cooling system. Total power consumption rate in mode “M” is 81,496 MW, where a humidifier and an exhaust air (EA) fan are not applied. Total power consumption rate in mode “M” is used as the baseline for comparing with the other operating modes. Operating mode “W M” is the case that the water-side economizer cycle is applied with the aid of mechanical cooling. In this mode, the water-side economizer cycle is operated when the wetbulb temperature of outside air is lower than 3 ◦ C (see Table 5). For

J.-Y. Kim et al. / Energy and Buildings 138 (2017) 257–270

263

Fig. 7. Comparison of the real time cooling energy consumption between experiment and simulation.

Table 5 Specifications of mechanical cooling system. Equipment

Quantity

Capacity

Power

Control mode

Chillers Cooling towers Chilled water pumps Condenser water pumps CRAH unit Exhaust fan Humidifier (mist spray type)

8 8 8 8 324 81 324

1,300 USRT 1,400 USRT 9,400 LPM 15,600 LPM 105 kW 912 CMH 113 kg/h

765 kW 37 × 3 kW 110 kW 160 kW 7.5 kW 30 kW 1 kW

Multiple units control Multiple units control Full operation/OFF in air-side mode Multiple units control CAV VAV ON/OFF control

the conditions of outdoor air of wet-bulb temperature above 3 ◦ C, the traditional mechanical cooling system is operated. The total power consumption rates of mode “W M” is 71,857 MW which

shows a reduction of 11.8% comparing with the basic operating mode “M”. Power consumption rates of cooling tower and condenser water pump increase, but those of the chiller decrease

264

J.-Y. Kim et al. / Energy and Buildings 138 (2017) 257–270

Fig. 8. Comparison of the cooling energy consumption rates between experiment and simulation.

Fig. 9. Weather conditions in Seoul, Korea.

J.-Y. Kim et al. / Energy and Buildings 138 (2017) 257–270 Table 6 Operating conditions of the MOA cooling system (when LW temp.†1 is 5 ◦ C and SA temp. †2 is 13 ◦ C). Operating mode

Outdoor air conditions

W WA A1

WB temp.†3 ≤ 3 ◦ C 3 ◦ C < WB temp.†3 ≤ 7 ◦ C Enthalpy ≤ 34.3 kJ/kg, and DB temp.†4 ≤ 13 ◦ C, and DP†5 ≤ 10.5 ◦ C 34.3 kJ/kg < Enthalpy ≤ 43.8 kJ/kg, and DB temp.†4 ≤ 21.5 ◦ C, and DP†5 ≤ 10.5 ◦ C 43.8 kJ/kg < Enthalpy, or 21.5 ◦ C < DB temp.†4 , or 10.5 ◦ C < DP†5

A2 M †1 †2 †3 †4 †5

LW temp.: Leaving water temperature of chilled water. SA temp.: Supply air temperature from CRAH unit. WB temp.: Wet-bulb temperature. DB temp.: Dry-bulb temperature. DP: Dew-point temperature.

substantially. Operating time of the water-side economizer cycle is 2,879 h. Outdoor air is not introduced in this mode, and the humidifier is also not needed. Operating mode “WA M” is a joint mode between water-side economizer and air-side economizer. In this mode, the water-side economizer cycle is the priority operating mode, which is aided by the air-side economizer only if the cooling capacity by the water-side economizer is not sufficient. That is, when the wetbulb temperature of outside air is lower than 3 ◦ C, the water-side economizer is operated in the same way as in operating mode “W”. When the wet-bulb temperature is above 3 ◦ C and lower than 7 ◦ C, the water-side economizer is fully operated and air-side economizer is operated to augment when necessary. For the outdoor air conditions beyond these conditions, the traditional mechanical cooling system is operated and both economizer cycles are shut off. The outdoor air volume introduced by the air-side economizers is controlled by regulating the OA damper in order to maintain the supply air temperature at 13 ◦ C. The total power consumption rates of operating mode “WA M” comes to 69,285 MW which is a 15.0% reduction compared with the basic mode “M”. In comparing with the operating mode “W M”, the power consumption of cooling tower and condenser water pump slightly increase, however, reduction rates of chiller more than compensate this. In this operating mode, the humidifier and EA fan are needed even though their power consumption rates are low at 280 MW and at 613 MW respectively. Operating time of economizer is 3,813 h which is 934 h longer than operating mode “W M”, and introduced outdoor air volume is 3,171 × 106 CMH. The humidifier operating time to keep the humidity level within the ASHRAE recommended range is 34.8

265

days, but its continuous operating time is relatively short, in this case 2.0 days. Humidifier operating time is a very critical point in a data center cooling system because its energy consumption rate can vary significantly by its type, e.g. the electric heater type is about 90 times of the mist spray type applied in this study. Operating mode “A1 M” is the case of utilizing air-side economizer cycles. In this mode, the air-side economizer is operated when the enthalpy of the outside air is lower than 34.3 kJ/kg, drybulb temperature lower than 13 ◦ C, and dew-point temperature is lower than 15 ◦ C. Introduced outdoor air volume is controlled by the OA damper as to set the supply air temperature at 13 ◦ C. For the conditions of outside air beyond these conditions, the mechanical cooling system is operated. The total power consumption rates reduce to 57,098 MW which is a 29.9% reduction compared with the basic operating mode “M”. Power consumption of cooling tower, condenser water pump, chiller and chilled water pump decrease largely. Humidifier and EA fan are needed in this mode, and their power consumption rates are 1,305 MW and 2,672 MW respectively. Operating time of air-side economizer is 4,376 h. Introduced outdoor air volume in this mode “A1 M” is 15,074 × 106 CMH. By this, humidifier operating time increases largely to 161.8 days, and its continuous operating time is 75.5 days. Operating mode “A2 M” is the case of utilizing air-side economizer cycles which is almost same with mode “A1 M”. The difference between these is the available outdoor air conditions. In addition to the outdoor air conditions for mode “A1 M”, when the outdoor air enthalpy is higher than the supply air enthalpy but lower than the return air enthalpy, 100% outdoor air is still brought in, but the chiller is operated to lower its temperature [20]. This additional conditions are that the outside air enthalpy is between 34.3 kJ/kg and 43.8 kJ/kg, the dry-bulb temperature is lower than 21.5 ◦ C, and the dew-point temperature is lower than 15 ◦ C. The total power consumption rates in this operating mode “A2 M” drop to 54,665 MW which shows a reduction of 32.9% compared with the basic mode “M”. In comparing with operating mode “A1 M”, power consumption rates of the cooling tower, condenser water pump and chiller become lower, however, those of humidifier and EA fan increase. Operating time of the air-side economizer is 5,081 h which is 705 h longer than operating mode “A1 M”. By this, the introduced outdoor air volume in this mode “A2 M” increases substantially, to 20,306 × 106 CMH, but the humidifier operating time increases slightly, to 169.1 days. These occur because humidity levels of outdoor air in the range between 34.3 kJ/kg and 43.8 kJ/kg are almost high enough. Operating mode “MOA” is the case composed of all operating modes in Fig. 4. In this study, the MOA cooling system is divided

Table 7 Annual energy conservation effects by economizer cycles and MOA cooling system. (L.W. temp. is 5 ◦ C and S.A. temp. 13 ◦ C). Power consump. rates (MW)

C.T. Condenser W.P. Chiller Chilled W.P. CRAH Humidifier EA fan Total (MW) Reduction rates (%) Economizer operation time (hours) OA volume (×106 CMH) Humidifier operation time (days) a b

Operating modes of economizer cycles M

WM

WA M

A1 M

A2 M

MOA1

2,287 7,343 42,870 7,709 21,287 0 0 81,496 0 0 (8,760)a 0 0.0 (0.0)b

4,300 9,186 29,374 7,709 21,287 0 0 71,857 11.8 2,879 (5,881)a 0 0.0 (0.0)b

4,842 9,624 24,931 7,709 21,287 280 613 69,285 15.0 3,813 (4,947)a 3,171 34.8 (2.0)b

1,388 4,287 22,303 3,861 21,287 1,305 2,672 57,098 29.9 4,376 (4,384)a 15,074 161.8 (75.5)b

1,200 3,638 18,933 3,858 21,287 1,364 4,386 54,665 32.9 5,081 (3,679)a 20,306 169.1 (94.4)b

4,703 9,147 22,303 7,214 21,287 331 1,526 66,510 18.4 4,376 (4,384)a 6,789 41.0 (2.0)b

Operation time of the mechanical cooling system. Continuous operation time of the humidifier to keep the humidity level within the ASHRAE recommended range [34].

MOA2 4,515 8,501 18,933 7,214 21,287 389 3,239 64,077 21.4 5,081 (3,679)a 11,996 48.2 (3.1)b

266

J.-Y. Kim et al. / Energy and Buildings 138 (2017) 257–270

Fig. 10. Monthly power consumption rates in MOA2 cooling system (LW temp. is 5 ◦ C and SA temp. 13 ◦ C).

into two types, that is, “MOA1” and “MOA2” for the comparison of their effects. Operating mode “MOA1” is composed of operating mode “W”, “WA”, “A1” and “M”. And operating mode “MOA2” has an additional mode “A2”. In operating mode “MOA1”, the total power consumption is 66,510 MW which is 18.4% reduction compared with the basic mode “M”. In comparing with the operating mode “A1 M”, the power consumption of the chiller are the same because the outdoor air conditions for chiller operation are identical. Power consumption rates of the humidifier and EA fan decrease, however, those of cooling tower, condenser water pump and chilled water pump increase significantly. In this operating mode “MOA1”, introduced outdoor air volume decreases a lot to 6,789 × 106 CMH which is 45% of operating mode “A1 M”. Because of this, humidifier operating time to maintain the humidity level within the ASHRAE recommended range also decreases a lot, to 41.0 days which is 25% of operating mode “A1 M”, and continuous operating time of humidifier is only 2.0 days. In operating mode “MOA2”, the total power consumption rates decrease to 64,077 MW which shows a reduction of 21.4% compared with the basic mode “M”. In comparing with the operating mode “A2 M”, power consumption of chiller is unchanged, and that of humidifier and EA fan decrease, but power consumptions of cooling tower, condenser water pump and chilled water pump increase by a lot. In this operating mode “MOA2”, introduced outdoor air volume is 11,996 × 106 CMH that is 59% of operating mode “A2 M”. Humidifier operating time is 48.2 days which is 29% of operating mode “A2 M”, and its continuous operating time decreases largely, to 3.1 days. For a more detailed analysis, data of each operating mode of “MOA2” cooling system are arranged in Table 8. The chiller is shut off in operating modes “W”, “WA”, and “A1” because cooling loads can be removed by economizers alone. Operating mode “W” is the mode that water-side economizer cycles deliver as much as a 11.8% power reduction, and its operating time is 2,879 h which is the longest among the operating time of the “MOA2” cooling system. It does not need to introduce outdoor air or apply a humidifier. Operating mode “WA” reduce the power consumption as much as 3.2%, and its operating time is 934 h. However it needs to introduce outdoor air for free cooling in the amount of 3,171 × 106 CMH, and, for this, needs the humidifier for 34.8 days. Operating mode “A1” and “A2” contribute to the power reduction at 3.4% and at 3.0% in each. And they need to introduce outdoor air as much as 3,618 × 106 CMH and as 5,207 × 106 CMH in each. Their humidifier operating time is 6.3 days and 7.2 days respectively. In the “MOA2” cooling system, operating time of the economizers is 5,081 h. The total operating time of the humidifier is 48.2 days, and, furthermore, its continuous operating time is quite short as of 3.1 days. Even though the

humidifier is shut off during certain times, the humidity level is maintained within the allowable range suggested by ASHRAE [19]. Considering that the continuous operating time of humidifier to keep the humidity level within the ASHRAE recommended range [34] is very short, it can be reasoned that it is possible to get rid of a humidifier in the “MOA2” cooling system. The main difference between “MOA1” and “MOA2” cooling in Table 8 is operating mode “A2”. Operating mode “A2” is changed to operating mode “M” in “MOA1” cooling. Power consumption in the “MOA1” system increases as much as 3.0% because of the contribution by operating mode “A2”. However, introduced outdoor air volume decreases by 43% compared to “MOA2” cooling. Fig. 10 shows monthly power consumption rates in the MOA2 cooling system when outlet water temperature of the chiller is at 5 ◦ C and supply air temperature of CRAH is set to 13 ◦ C. From January to March, operating mode “W” is the prominent mode, and on April, all operating modes occur. From January to April, the monthly power consumption of the MOA2 system is about 63% of the mechanical only system that is denoted as “M only WR”. In May, monthly power consumption rates of MOA2 cooling is 87% of the mechanical only cooling system, and operating mode “W” does not occur. From June to August, there is only operating mode “M”, and monthly power consumption rates are the same as for the mechanical system only. In September, operating mode “M” is the prominent mode, and power consumption rates of MOA2 cooling is about 98% of the mechanical only system. In October, operating mode “M” is the prominent mode, and operating modes “WA”, “A1” and “A2” occur. In November, operating modes “WA” and “W” are the prominent modes, and operating modes “A1” and “A2” occur. In December, operating mode “W” is the prominent mode, and operating mode “WA” occurs. From October to December, the monthly power consumption of the MOA2 system is about 66% of the mechanical only cooling system. In case of the MOA1 cooling system, operating mode “A2” is changed to operating mode “M”, and, thus, monthly power consumption rates are increased analogously. 5.2. Effects of MOA cooling system combined with the rack enclosure The MOA cooling system combined with the rack enclosure (See 3, “Rack enclosure”) is expected to show an improved energy performance. Fig. 11 shows the concept of a rack enclosure type that is under development. By combining the MOA cooling system and rack enclosure, in this study it is assumed that there is no air mix between cold supply air and hot exhaust air, and thus, it is possible to increase the supply air temperature to 25 ◦ C. This supply

J.-Y. Kim et al. / Energy and Buildings 138 (2017) 257–270

267

Table 8 Annual energy conservation effects in each modes of MOA cooling system. (L.W. temp. is 5 ◦ C and S.A. temp. 13 ◦ C). Power consump. rates (MW)

C.T. Condenser W.P. Chiller Chilled W.P. CRAH Humidifier EA fan Total (MW) Reduction rates (%) Operation time (hours) OA volume (×106 CMH) Humidifier operation time (days) a

Operating modes of MOA cooling system W

WA

A1

A2

M

MOA2

2,549 3,685 0 2,533 6,996 0 0 15,763 11.8 2,879 0 0.0 (0.0)a

767 1,178 0 822 2,270 280 613 5,931 3.2 934 3,171 34.8 (2.0)a

0 0 0 0 1,367 51 913 2,330 3.4 562 3,618 6.3 (0.5)a

10 40 172 620 1,713 58 1,713 4,327 3.0 705 5,207 7.2 (0.5)a

1,189 3,598 18,760 3,238 8,941 0 0 35,726 0 3,679 0 0.0 (0.0)a

4,515 8,501 18,933 7,214 21,287 388 3,239 64,077 21.4 8,7608,760 11,996 48.2 (3.1)a

Continuous operation time of the humidifier to keep the humidity level within the ASHRAE recommended range [34]. Table 9 Operating conditions of MOA cooling system combined with the rack enclosures (when LW temp.†1 is 17 ◦ C and SA temp. †2 is 25 ◦ C). Operating mode

Outdoor air conditions

W WA A1

WB temp. †3 ≤15 ◦ C 15 ◦ C < WB temp. †3 ≤19 ◦ C, and DP†4 ≤15 ◦ C Enthalpy ≤ 52.1 kJ/kg, and DB temp. †5 ≤25 ◦ C, and DP†4 ≤15 ◦ C 52.1 kJ/kg < Enthalpy ≤ 61.8 kJ/kg, and 25 ◦ C < DB temp. †5 ≤33.5 ◦ C, and DP†4 ≤ 15 ◦ C 61.8 kJ/kg < Enthalpy, or 33.5 ◦ C < DB temp. †5 , or 15 ◦ C < DP†4

A2 M †1 †2 †3 †4 †5

Fig. 11. Rack enclosure under the development.

air temperature is derived from the considerations that the upper limit temperature in a class 1 data center is 27 ◦ C [35], and it is desirable to take 2 ◦ C lower as a safety factor. Table 9 shows the operating conditions of the MOA cooling system combined with the rack enclosure. It is assumed that the outlet water temperature is at 17 ◦ C and the supply air temperature at 25 ◦ C. For combining with the rack enclosure, an additional rack fan is needed for the exhaust air. Table 10 shows the simulation results in each operating mode combined with the rack enclosure. Mode “M” is the traditional mechanical cooling combined with the rack enclosure. Total power consumption in mode “M” is 65,409 MW which shows a 19.7% power reduction compared with the basic operating mode “M” in Table 7. This reduction effect is the result of the improved COP of

LW temp.: Leaving water temperature of chilled water. SA temp.: Supply air temperature from CRAH unit. WB temp.: Wet-bulb temperature. DP: Dew-point temperature. DB temp.: Dry-bulb temperature.

the chiller by increasing the outlet water temperature from 5 ◦ C to 17 ◦ C. A humidifier and an EA fan are not needed in this mode. Operating mode “W M” is the mode of water-side economizer cycle with the aid of mechanical cooling. In this mode, the waterside economizer cycle is operated when the wet-bulb temperature of the outside air is lower than 15 ◦ C. For the conditions of outdoor air with wet-bulb temperature above 15 ◦ C, the mechanical cooling system is operated. The total power consumption of mode “W M” is 55,774 MW which is 31.6% power reduction compared with the basic operating mode “M” in Table 6. Operating time of the water-side economizer cycle is 5,788 h which is more than twice that of operating mode “W M” in Table 7. This is the result of the increased time of useable available outdoor air conditions by increasing the outlet water temperature. A humidifier and an EA fan are not needed in this mode. In operating mode “WA M”, the total power consumption falls to 55,350 MW which is 32.1% of the reduction compared with the basic mode “M”. In this operating mode, the humidifier is not needed, but the EA fan is needed and its power consumption is 226 MW. Operating time of economizers is 6,386 h which is 67% longer than operating mode “WA M” in Table 6. However, introduced outdoor air volume is 1,562 × 106 CMH which is 51% smaller than in operating mode “WA M” in Table 7. Operating mode “A1 M” is the mode of air-side economizer cycles. In this mode, air-side economizer cycles are operated when the enthalpy of the outside air is lower than 52.1 kJ/kg, dew-point temperature lower than 15 ◦ C, and dry-bulb temperature is lower than 25 ◦ C which is the supply air set temperature. The total power consumption reduces to 40,919 MW which is 49.8% reduction compared with basic mode “M”. Operating time of air-side economizer advances to 6,313 h which is 44% longer than operating mode

268

J.-Y. Kim et al. / Energy and Buildings 138 (2017) 257–270

Table 10 Annual energy conservation effects by economizer cycles and MOA cooling system. (L.W. temp. is 17 ◦ C and S.A. temp. 25 ◦ C). Power consump. rates (MW)

C.T. Condenser W.P. Chiller Chilled W.P. CRAH Humidifier EA fan Rack fan Total (MW) Reduction rates (%) Economizer operation time (hours) OA volume (×106 CMH) Humidifier operation time (days) a b

Operating modes of economizer cycles M

WM

WA M

A1 M

A2 M

MOA1

2,143 6,677 23,959 7,709 21,287 0 0 3,633 65,409 19.7 0 (8,760)a 0 0.0 (0.0)b

4,732 9,904 8,510 7,709 21,287 0 0 3,633 55,774 31.6 5,788 (2,972)a 0 0.0 (0.0)b

5,555 10,146 6,794 7,709 21,287 0 226 3,633 55,350 32.1 6,386 (2,374)a 1,562 0.0 (0.0)b

780 2,109 7,007 2,153 21,287 1,299 2,652 3,633 40,919 49.8 6,313 (2,447)a 17,539 161.0 (94.4)b

756 2,041 6,791 2,153 21,287 1,299 2,831 3,633 40,791 49.9 6,387 (2,373)a 18,085 161.0 (94.4)b

5,102 10,146 6,794 7,709 21,287 0 226 3,633 54,897 32.6 6,313 (2,447)a 1,562 0.0 (0.0)b

MOA2 5,102 10,145 6,791 7,709 21,287 0 229 3,633 54,895 32.6 6,387 (2,373)a 1,571 0.0 (0.0)b

Operation time of the mechanical cooling system. Continuous operation time of the humidifier to keep the humidity level within the ASHRAE recommended range [34].

Table 11 Annual energy conservation effects by MOA cooling system. (L.W. temp. is 17 ◦ C and S.A. temp. 25 ◦ C). Power consump.

rates (MW)

C.T. Condenser W.P. Chiller Chilled W.P. CRAH Humidifier EA fan Rack fan Total (MW) Reduction rates (%) Operation time (hours) OA volume (×106 CMH) Humidifier operation time (days)

Operating modes of MOA cooling system W

WA

A1

A2

M

4,267 7,339 0 5,094 14,065 0 0 2,400 32,713 24.2 5,788 0 0.0

525 763 0 526 1,452 0 235 248 3,747 2.6 598 1,562 0.0

0 0 0 0 0 0 0 0 0 0.0 0 0 0.0

0 0 0 0 3 0 3 0 6 0 1 9 0.0

756 2,039 6,784 2,088 5,767 0 0 984 8,968 5.9 2,373 0 0.0

“A1 M” in Table 6. Humidifier and exhaust air fan are needed in this mode, and their power consumption rates are 1,299 MW and 2,652 MW respectively. The introduced outdoor air volume is 17,539 × 106 CMH which is 16% larger than operating mode “A1 M” in Table 7, but humidifier operating time is almost the same, 161.0 days. In operating mode “A2 M”, in addition to operating mode “A1 M”, the air-side economizer cycle is operated when the outside air enthalpy is in between 52.1 kJ/kg and 61.8 kJ/kg, dew-point temperature is lower than 15 ◦ C, and the dry-bulb temperature is lower than 33.5 ◦ C which is the return air temperature. The total power consumption drops to 40,791 MW which shows the reduction rates of 49.9% comparing with the basic mode “M” in Table 7. Operating time of air-side economizer advances to 6,387 h which is 26% longer than operating mode “A2 M” in Table 7, but introduced outdoor air volume is 11% smaller as of 18,085 × 106 CMH. Humidifier operating time is the same with the mode “A1 M” as of 161.0 days, because a humidifier is not needed in the range of outdoor air conditions added for operating mode “A2 M”. In operating mode “MOA1” combined with the rack enclosure, the total power consumption is 54,897 MW which is 32.6% reduction compared with the basic mode “M” in Table 7. Introduced outdoor air volume is 1,562 × 106 CMH which is 91% smaller than operating mode “A1 M”, and a humidifier is not needed in this mode. In operating mode “MOA2” combined with the rack enclosure, the total power consumption rates decrease to 54,896 MW which is almost the same with operating mode “MOA1”. Introduced outdoor

MOA2 5,547 10,141 6,784 7,709 21,287 0 238 3,633 54,894 32.6 8,760 1,607 0.0

air volume is 1,571 × 106 CMH that is 91% smaller than operating mode “A2 M”. A humidifier is not needed in this operating mode, too. For more detailed analysis, data of each operating mode in “MOA2” cooling system are arranged in Table 11. The chiller is shut off in operating modes “W” and “WA” because cooling loads can be removed by economizers only. Operating mode “W” contributes as much as 24.2% power reduction, and its operating time is 5,788 h which is more than twice of operating mode “W” in Table 7 which is without the rack enclosure. In operating mode “WA”, the power consumption reduces by 2.6%, and the operating time of the economizers is 598 h which is 36% shorter than in operating mode “WA” in Table 8. Operating mode “WA” needs to introduce outdoor air for free cooling in the amount of 1,562 × 106 CMH which is 51% smaller than mode “WA” in Table 8. However, these modes do not need a humidifier because a humidity level of outdoor air is high enough. Operating mode “A1” does not occur in this “MOA2” cooling system. Operating mode “A2” occurs only for one hour, and the total power consumption is very small: 6 MW. In the “MOA2” cooling system combined with the rack enclosure, operating time of economizers is 6,387 h which is 8% longer than the former “MOA2” cooling system in Table 8 which is without the rack enclosure. In the “MOA2” cooling system combined with the rack enclosure, humidifier is not needed at all. In case of “MOA1” cooling combined with the rack enclosure, operating mode “A2” is changed to operating mode “M”. However, the results are almost unchanged because the operating time of operating mode “A2” is only one hour.

J.-Y. Kim et al. / Energy and Buildings 138 (2017) 257–270

269

Fig. 12. Monthly power consumption rates in MOA2 cooling system (LW temp. is 17 ◦ C and SA temp. 25 ◦ C).

Fig. 12 shows the monthly power consumption rates of MOA2 cooling system combined with the rack enclosure when leaving water temperature of chiller is maintained at 17 ◦ C and supply air temperature of CRAH at 25 ◦ C. Mechanical only cooling under these operating conditions is denoted as “M only R” whereas mechanical only cooling under the former conditions, leaving water temperature of chiller at 5 ◦ C and supply air temperature at 13 ◦ C, is denoted as “M only WR”. Power consumption of M only R is about 20% lower than M only WR for every month because of the improved COP of the chiller. From January to May, operating mode “W” is the prominent mode, and monthly power consumption of the MOA2 system is about 77% of “M only R”. In June, operating modes “M” is the prominent modes, and monthly power consumption of the MOA2 system is 94% of “M only R”. From July to August, only operating mode “M” occurs, and monthly power consumption rates are the same as for “M only R”. In September, operating modes “M” is the prominent modes, and power consumption of the MOA2 cooling system is about 90% of “M only R”. From October to December, operating mode “W” is the prominent mode, and monthly power consumption of MOA2 cooling is about 78% of “M only R”. 6. Conclusions In this study, a safer and energy conserving system for data centers, based on a multi-stage outdoor air enabled (MOA) cooling strategy was established. The effects of the MOA cooling system were studied through a validated simulation method which leads to the following findings: • Cooling energy conservation of 18.4% was established for the MOA1 cooling system and 21.4% for the MOA2 cooling system compared with traditional mechanical cooling systems. • Supplied outdoor air volume in the MOA1 system drops to 45% of air-side economizer cycles; for the MOA2 system it drops to 59%. • The total operating time of the humidifier in MOA1 decreases to 41.0 days which is 25% of the normal air-side economizer, and the continuous operating time is only 2.0 days; in MOA2 the total operating time decreases to 48.2 days which is 29% of the air-side economizer and the continuous operating time is 3.1 days. • The MOA cooling system combined with a rack enclosure shows a power reduction of 32.6% in both of MOA1 and MOA2 compared with the standard mechanical cooling system. • Supplied outdoor air volume in both of MOA1 and MOA2 combined with rack enclosure decreases to 9% of the air-side economizer combined with the rack enclosure. • A humidifier is not needed in both MOA1 and MOA2 cooling systems combined with the rack enclosure.

• The results show a significant reduction in operating time of the humidifier, which leads to the reasoning that it should be possible to remove the humidifier in the MOA cooling system altogether; this conclusion is true irrespective of the combination with rack enclosure. Acknowledgement This research was supported by a grant (13CHUD-C06196702) from Construction & Transportation Technology Advancement Research Program funded by Ministry of Land, Infrastructure and Transport of Korean government. References [1] Daniel Hallett, Christian Paunon, Control strategies for data centers – trends around the globe, 2014, ASHRAE Winter Conference, NY-14-C094 (2014). [2] Michael Rutberg, Alissa Cooperman, Antonio Bouza, Data center cooling, ASHRAE J. (2013) 82–86. [3] C. Belady, Projecting Annual New Datacenter Construction Market Size, Microsoft, 2011 http://tinyurl.com/kb4o6ye. [4] ASHRAE, ASHRAE Datacom Series: Best Practices for Datacom Facility Energy Efficiency, 2nd ed., 2009, pp. 17–18. [5] LBNL, Benchmarking: Data Centers-charts, Lawrence Berkeley National Lab., 2007 http://hightech.lbl.gov/benchmarking-dc-charts.html. [6] Steven Pelley David Meisner Thomas F. Wenisch. James W. VanGilder, Understanding and Abstracting Total Data Center Power, In: Workshop on energy-efficient design. [7] Jinkyun Cho, Taesub Lim, Byungseon Sean Kim, Viability of datacenter cooling systems for energy efficiency in temperature or subtropical regions: case study, Energy Build. 55 (2012) 189–197. [8] Hao Tiana, Zhiguang Heb, Zhen Lib, A combined cooling solution for high heat density data centers using multi-stage heat pipe loops, Energy Build. 94 (2015) 177–188. [9] D. Christy Sujatha, Satheesh Abimannan, Energy efficient free cooling system for data centers, Third IEEE International Conference on Cloud Computing Technology and Science, 2011 (2011) 646–651. [10] Aayush Agrawal 1, Mayank Khichar 1, Sanjeev Jain, Transient simulation of wet cooling strategies for a data center in worldwide climate zones, Energy Build. 127 (2016) 352–359. [11] Bruce A. Hellmer, Consumption analysis of telco and data center cooling and humidification options, ASHRAE Trans. 116 (Part 1) (2010) 118–133. [12] Jayantha Siriwardana, Saliya Jayasekara, Saman K. Halgamuge, Potential of air-side economizers for data center cooling: a case study for key Australian cities, Appl. Energy 104 (2013) 207–219. [13] Kuei-Peng Lee, Hsiang-Lun Chen, Analysis of energy saving potential of air-side free cooling for datacenters in worldwide climate zones, Energy Build. 64 (2013) 103–112. [14] Sang-Woo Ham, Jun-Seok Park, Jae-Weon Jeong, Optimum supply air temperature ranges of various air-side economizers in a modular data center, Appl. Therm. Eng. 77 (2015) 163–179. [15] Sang-Woo Ham, Min-Hwi Kim, Byung-Nam Choi, Jae-Weon Jeong, Energy saving potential of various air-side economizers in a modular data center, Appl. Energ. 138 (2015) 258–275. [16] Sang-Woo Ham, Jae-Weon Jeong, Impact of aisle containment on energy performance of a data center when using an integrated water-side economizer, Appl. Therm. Eng. 105 (2016) 372–384.

270

J.-Y. Kim et al. / Energy and Buildings 138 (2017) 257–270

[17] Milnes P. David, Madhusudan Iyengar, Pritish Parida, Robert Simons, Mark Schultz, Michael Gaynes, Roger Schmidt, Timothy Chainer, Experimental characterization of an energy efficient chiller-less data center test facility with warm water cooled servers, 28th IEEE SEMI-THERM Symposium (2012) 232–237. [18] Hiroshi Endo, Hiroyoshi Kodama, Hiroyuki Fukuda, Toshio Sugimoto, Takashi Horie, Masao Kondo, Effect of climatic conditions on energy consumption in direct fresh-air container data centers, Sustain. Comput.: Inf. Syst. 6 (2015) 17–25. [19] Baptiste Durand Estebe, Cedric Le Bot, Jean Nicolas Mancos, Eric Arquis, Simulation of a temperature adaptive control strategy for an IWSE economizer in a data center, Appl. Energy 134 (2014) 45–56. [20] Alex Myers, Vali Sorell, Integrated Waterside Economizer Operation for Data Center Central Chiller Plants, White Paper #14, Syska Hennessy Group, 2009. [21] Eduard Oró, Victor Depoorter, Noah Pflugradt, Jaume Salom, Overview of direct air free cooling and thermal energy storage potential energy savings in data centres, Appl. Therm. Eng. 85 (2015) 100–110. [22] Eduard Oró, Victor Depoorter, Albert Garcia, Jaume Salom, Energy efficiency and renewable energy integration in data centres Strategies and modelling review, Renew. Sustain. Energy Rev. 42 (2015) 429–445. [23] Ali Habibi Khalaj, Thomas Scherer, Saman K. Halgamuge, Energy, environmental and economical saving potential of data centers with various economizers across Australia, Appl. Energy 183 (2016) 1528–1549. [24] Jinkyun Cho, Joonyoung Yang, Changkeun Lee, Jinyoung Lee, Development of an energy evaluation and design tool for dedicated cooling systems of data centers: sensing data center cooling energy efficiency, Energy Build. 96 (2015) 357–372.

[25] Jinkyun Cho, Yundeok Kim, Improving energy efficiency of dedicated cooling system and its contribution towards meeting an energy-optimized data center, Appl. Energy 165 (2016) 967–982. [26] ASHRAE, ASHRAE Datacom Series: Best Practices for Datacom Facility Energy Efficiency, 2nd ed., 2009, pp. 53–66. [27] LBNL, Data Center Economizer Contamination and Humidity Study, Lawrence Berkeley National Lab, 2007. [28] Jinkyun Cho, Byungseon Sean Kim, Evaluation of air management system’s thermal performance for superior cooling efficiency in high-density data centers, Energy Build. 43 (2011) 2145–2155. [29] ASHRAE, Green Tips for Data Centers, 2010, p. 25. [30] Kishor Khankari, Rack enclosures, ASHRAE J. (August) (2009) 48–51. [31] ASHRAE, ASHRAE Datacom Series: Best Practices for Datacom Facility Energy Efficiency, 2nd ed., 2009, pp. 39–41. [32] ASHRAE, ASHRAE Datacom Series: Thermal Guidelines for Data Processing Environments, 2nd ed., 2009, pp. 25–29. [33] KMA, Home page of Korea Meteorological Administration. (http://web.kma. go.kr/child/normal/season.js). [34] ASHRAE, ASHRAE Datacom Series: Thermal Guidelines for Data Processing Environments, 2nd ed., 2009, pp. 9–13. [35] Walter T. Grondzik, Alison G. Kwok, Benjamin Stein, John S. Reynolds, Mechanical and Electrical Equipment for Buildings, 2010, p. 416.