Experimental Investigation of Heat Transfer and Flow Characteristics of Split Natural Cooling System for Data Center Based on Micro Heat Pipe Array

Abstract

This paper presents a new type of split natural cooling system that maximizes the use of natural cold energy to significantly reduce the power consumption of the air conditioning system in data centers. A split natural cooling system module, which consisted of indoor and outdoor heat exchanger based on micro heat pipe arrays connected by liquid circulation system, was selected for experimental research. The heat transfer process and flow characteristics were analyzed under different outdoor environment temperatures, air and water flow rates, and different ratios of heat transfer components (N) of indoor and outdoor heat exchangers. To improve the utilization of natural cold energy, two kinds of heat dissipation conditions, namely room and heat channel-based, were proposed. The indoor temperature of two conditions at 28 °C and 38 °C were simulated in the laboratory at constant temperature-humidity, respectively. Results indicated that the air flow rate had a greater influence on the heat transfer performance than the water flow rate. The pressure drop of the air and water sides was at a lower level, and the fitting curve of the pressure drop was obtained to provide a reference for the heat exchanger design and equipment selection. When the ratio of heat transfer components (N) of the indoor and outdoor heat exchanger was approximately 0.75, the split natural cooling system showed optimal comprehensive performance. Under heat channel-based conditions, the maximum heat transfer rate reached 12.4 kW, and the maximum energy efficiency ratio was 17.15; the maximum heat transfer rate and the maximum energy efficiency ratio increased by 42.5% and 22.64% compared with the room-based condition, respectively. The fitting curve of the energy efficiency ratio was calculated under different outdoor temperatures at two heat dissipation conditions.

1. Introduction

The increasing heat dissipation of equipment in a data center greatly increases the energy consumption of the air conditioning (AC) system [1,2,3,4]. The American Society of Heating, Refrigerating, and Air-Conditioning Engineers (ASHRAE) has proposed that the indoor temperature of medium-sized data centers should be controlled within 28 °C and the humidity should be maintained at about 40% [5]. The AC system of data centers works continuously throughout the year for 8760 h with huge energy consumption [6,7,8]. Thus, improving the cooling efficiency of data centers is important [9].

Technologies using natural cold energy in data centers reduce power consumption [10,11]. The application of a natural cooling energy in a data center is mainly divided into three categories [12,13], which are: (i) room-based heat dissipation for the overall environment throughout the space; (ii) heat channel-based heat dissipation between servers and cabinets; and (iii) chip and rack-based heat dissipation. The complicated third heat dissipation condition requires top design according to the structural characteristics of different chips [14]. The simple forms of room-based and heat channel-based heat dissipations are more suitable for energy saving in data centers.

As the simplest form of natural cold energy, the direct free air-cooling system can improve the utilization of natural cold energy effectively and reduce the comprehensive costs [15,16]. However, a data center is subjected to risks that include cleanliness, humidity, and the temperature variation of outdoor cold air [17]. According to Ham [18] and Siriwardana [19], indirect kinds of air-side economizers are used to introduce outside air with the desired air supply conditions by exploiting cool and dry climate conditions. The results show that the total cooling energy saved by the economizers increases by 42% compared to conventional cooling systems in data centers. However, these types are constrained by high outdoor temperature and humidity requirements of air flow; moreover, the economizers are too large to meet the requirements of the heat transfer area [20,21]. The air-cooled chillers are gradually replaced by water-cooled chillers due to their low efficiency, and a large redundancy is necessary to ensure the reliability and stability of data centers [22,23]. Other studies have proposed the combination of an open or closed cooling tower with different heat exchanger equipment on the indoor side [24,25]. The open cooling tower has the best direct cooling effect, but the system needs additional water. Moreover, when the outdoor environment temperature is low, the frozen problem is difficult to overcome, the heat transfer performance of a closed cooling tower is slightly insufficient, and the pipeline resistance is large. These factors waste resources and increase initial investment. Li et al. [26] have used natural cold energy to save energy according to the temperature variations of different seasons and have verified the energy saving effect of air conditioning systems in the climate of Northern China.

The multiple combinations of the economizer and water/air-cooled chiller system and the combined cooling heating and power system are proposed to use further energy [27,28,29]. The complexity of the above cooling technology needs intelligent variable frequency controls, which require the quantitative understanding of their economic benefits [30,31,32].

Therefore, natural cold energy application has become very important in the data center. To enhance the heat transfer capacity and the uniformity of the temperature distribution, numerous scholars have applied heat pipe cooling systems to data centers. Scholars [33,34] have proposed a split heat pipe heat exchanger system in which the mismatch degree is reduced by increasing the heat pipe series. In addition, the heat transfer efficiency of a three-stage heat pipe system is less than 65%. Han et al. [35] have developed an integrated system, which combines a split heat pipe with air conditioning. The energy efficiency ratio (EER) can reach 9.43 when the difference between the indoor and outdoor temperatures is 20 °C. Feng et al. [36] have proposed a pump-driven loop heat pipe (PLHP) system to overcome the resistance of refrigerant circulation for free cooling; the annual energy saving is about 30% and the payback period is about 4 years. Moreover, the cooling efficiency of the split loop heat pipe is evidently higher than mechanical refrigeration for cooling data center [37,38]. However, the above heat pipe system is directly affected by the mismatch between pipeline length and liquid filling rate, resulting in the impact on heat transfer capacity and system stability.

Zhao et al. [39] proposed a micro heat pipe array (MHPA) with flat shape, which enhances the heat transfer rate and increases the contact area [40,41,42]. Wang [43,44] has studied the heat transfer performance of heat exchangers with different kinds of fins and different arrangement modes. The results show that the number of rows on the windward side of the heat exchanger and the number of rows along the length direction have a great influence on the heat transfer performance and resistance performance, and the type of serrated fins show a better performance than the flat, perforated, corrugated, and louver types. Liang et al. [45] have improved the structure of an air–air heat exchanger based on MHPA with welded serrated fins to enhance the heat transfer—the highest heat transfer efficiency reaches 71%. According to these studies, a series of heat exchangers with MHPA as the core component exhibits a better performance.

This study proposes the application of a new type of split natural cooling system (SNCS) consisting of indoor and outdoor air–water heat exchangers (AWHE) to data centers by using natural cooling energy to reduce energy consumption. The SNCS has the following advantages:

(i)

On the air side of AWHE, the MHPA combined with serrated fins on its surface possesses a large convective heat transfer area. The problem of the small contact area between conventional round heat pipes and fins is solved. The air flow disturbance is enhanced and energy efficiency improves;

(ii)

On the water side of AWHE, the flat-plate appearance of a parallel flow tube (PFT) can increase the contact area more effectively with the MHPA. The convective heat transfer area is greatly increased by dozens of independent tiny porous channels to reduce the thermal resistance of heat transfer;

(iii)

The disassembly and assembly of the AWHE with a compact structure are simple, and the material of each part of the AWHE is made of light aluminum;

(iv)

The simple application of SNCS in different areas, which can choose water or glycol antifreeze instead of refrigerants as the circulating medium, is not limited by transmission distance. The water-cooled system is relatively flexible and reliable because it has no icing and a direct connection with the servers and electronic equipment.

According to the super thermal conductivity of the micro heat pipe array, the split natural cooling system can improve the indoor and outdoor heat transfer efficiency, improve the heat exchange capacity and the utilization time of natural cold energy, in order to minimize the energy consumption of air conditioning system.

The Enthalpy Potential Method Laboratory is used to simulate two heat dissipation conditions of indoor temperature at 28 °C and 38 °C. The heat transfer characteristics, pressure drop, power consumption, and EER of an SNCS module under different outdoor environment temperatures, air and water flow rates, and different ratios of heat transfer components (N) of indoor and outdoor MHPA–AWHEs, are studied. The comprehensive performance of the SNCS module provides a reference for its application in data centers.

2. Experimental Investigation

2.1. Split Natural Cooling System

The new type of split natural cooling system (SNCS) consists of indoor and outdoor MHPA–AWHEs, and each side is combined with the water-cooled circulation loop. It avoids the direct contact between indoor and outdoor air and overcomes the influence of cleanliness, humidity and other unstable factors of a data center. As shown in Figure 1, the hot air is sent to the indoor MHPA-AWHE through the centrifugal fan, and then the heat transfer to the outdoor MHPA-AWHE through the liquid circulation is finally sent back to the data center after cooling by the natural cooling system. The composition will be introduced and displayed in order from parts to the whole system.

2.2. MHPA–AWHE

The core heat transfer unit of AWHE is the MHPA. The MPHA consists of several capillary microgroove structures working independently. One or two damaged micro heat pipes will not affect the overall performance. The filling fluid is R141b, the filling rate is at about 20%. The working fluid by complex phase transformation in a vacuum state ensures the high performance of heat transfer and uniform temperature distribution, as shown in Figure 2. The width and length of the MHPA are 80 mm and 1000 mm and the thickness is only 3 mm. The serrated fins on the MHPA’s surface can increase the heat transfer area and enhance the air flow disturbance. The height and width of each serrated fin are 12 mm and 3 mm, and the thickness is 0.2 mm.

The parallel flow tube (PFT) is another heat transfer unit with 22 tiny porous channels and is flat and easy to fit with the MHPA. The width and height of each channel are 4.5 mm and 4 mm, and the thickness is 1.2 mm, as shown in Figure 3a. The width and length of PFT are 820 and 120 mm. The microgroove structure of the inner wall can enlarge the convective heat transfer area. A heat transfer component of MHPA–AWHE is composed of 1 PFT and 10 MHPAs with serrated fins, as shown in Figure 3b.

The indoor and outdoor MHPA–AWHEs have the same principle with different forms. According to theoretical calculation, the experimental indoor MHPA–AWHE consists of 12 rows of heat exchanger components. Each row of the heat exchanger component is fixed by a comb-like roof that is easy to install, as shown in Figure 4. The MHPA with a serrated fin on the air side is the evaporation section, and the MHPA pasted with PFT by silica gel on the water side is the condensation section. In contrast, the outdoor MHPA–AWHE on the air side is the condensation section, and the evaporation section of the MHPA is on the water side. The three forms of outdoor MHPA–AWHE are 12, 16, and 20 rows of heat exchanger components. The three types of SNCS during the experiment are studied, and the different types with different N are shown in

Table 1. Three types of SNCS during the experiment.

Ratio of Heat Transfer Components (N)	Heat Transfer Components of Indoor MHPA–AWHE	Heat Transfer Components of Outdoor MHPA–AWHE
N = 1	12	12
N = 0.75	12	16
N = 0.6	12	20

.

2.3. The Form of Its Application

The new type of split natural cooling system (SNCS) consists of indoor and outdoor MHPA–AWHEs, and each side is combined with the water-cooled circulation loop. According to the super thermal conductivity of MHPA, the SNCS can improve the heat exchange capacity and the utilization time of natural cold energy; it only needs little energy consumption of fans and pumps to reduce the consumption of the air conditioning compressor in order to minimize the energy consumption of air conditioning system.

The SNCS is mainly used for energy saving in small and medium-sized data centers, and two forms of its application are presented: (i) The room-based form of heat source temperature at 28 °C is relatively a simple form throughout the space, as shown in Figure 5a; (ii) A heat channel-based form of heat source temperature at 38 °C between server cabinets is shown in Figure 5b.

In winter and transition seasons, the SNCS transmits heat energy from indoors to outdoors. Cooled air is delivered back to AC to reduce the working time and power consumption of AC. Then, the air is sent to each server by the underfloor ventilation systems. Several advantages of the heat channel-based form over the room-based form include: (i) a widely used temperature difference between the indoor side and the outdoor side; (ii) a higher heat transfer rate and energy efficiency ratio (EER); and (iii) fewer indoor MHPA–AWHEs can meet the cooling load of the data center.

2.4. Experimental System

The experimental SNCS consists of four sections: indoor and outdoor heat transfer, liquid circulation, and a data acquisition section, as shown in Figure 6.

The indoor and outdoor heat transfer sections have the same experimental equipment: MHPA–AWHEs, frequency conversion centrifugal fans, air handling units, and wind resistance monitors. The closed water-cooled circulation carries heat energy from indoor to outdoor MHPA–AWHE. The differential pressure transmitters are used to measure the pressure drop on the air side and the water side. The acquisition equipment records data every 10 s. The power monitor measures the power consumption of the water pump and fans. The temperature is measured by a thermocouple and thermal resistance. Four thermal resistors are used to monitor the water temperature of the inlet and outlet from inside and outside of the room The test equipment and parameters are showed in

Table 2. Information of experimental instruments.

Instrument	Model	Working Scale	Accuracy	Number
Frequency conversion pump	CHL2–20	6–15 m	–	1
Centrifugal fan	Popula–9#1.5KW	0–3200 m3/h	–	1
Water flow meter	MFM–15	0–3 m3/h	0.5%	1
Data logger	Agilent 34972A	0–300 V	–	1
Thermal resistor	PT100	0–150 °C	±0.1 °C	4
Differential pressure Transmitter of air side	3351DP5SM3	0–30 kPa	0.5%	1
Differential pressure Transmitter of water side	3351DP5SM5	0–50 kPa	0.5%	1
Power monitor	(HYELEC)HY–001	0–10 A	1%	1
Expansion tank	AP5C/5L	8 Bar	–	1

.

2.5. Experimental Method

The different air flow rates on the air side of indoor and outdoor MHPA-AWHE are 1000, 1500, 2000, 2500, and 3000 m³/h, respectively; the air flow rate of indoor and outdoor MHPA–AWHEs remain the same. The different flow rates on the water side are 400, 600, 800, 1000, and 1200 L/h, respectively. The outdoor temperature is set at −15 °C, −10 °C, −5 °C, 0 °C, 5 °C, 10 °C, and 15 °C, respectively. During the experiment, comprehensive performances are analyzed. The thermal performance, pressure drop, and EER are analyzed under the two different heat dissipation conditions and the SNCS module at different N.

2.6. Evaluation Index of the SNCS

2.6.1. Heat Transfer Rate and Loss Percentage

The heat transfer rate determines the heat exchange capacity of the heat exchanger. This rate includes indoor and outdoor air- and water-side. The heat loss percentage Δβ is used to measure the level of heat loss and heat balance of each side. The expressions are as follows:

Heat transfer rate of the air and water sides of indoor MHPA–AWHE:

$$Q_{IN,a}=c_{IN,a}{\rho }_{IN,a}{q}_{V,IN,a}(T_{IN,a,i}-T_{IN,a,o})$$

(1)

$$Q_{IN,w}=c_{IN,w}{\rho }_{IN,w}{q}_{V,IN,w}(T_{IN,w,o}-T_{IN,w,i}).$$

(2)

Heat transfer rate of the air and water sides of outdoor MHPA–AWHE:

$$Q_{OUT,a}=c_{OUT,a}{\rho }_{OUT,a}{q}_{V,OUT,a}(T_{OUT,a,o}-T_{OUT,a,i})$$

(3)

$$Q_{OUT,w}=c_{OUT,w}{\rho }_{OUT,w}{q}_{V,OUT,w}(T_{OUT,w,i}-T_{OUT,w,o}).$$

(4)

Heat loss percentage:

$$△\beta =\frac{Q-Q_{AVE}}{Q_{AVE}},$$

(5)

where Q_AVE is the average value of heat transfer rates of the indoor and outdoor air and water side.

2.6.2. Thermal Resistance of MHPA–AWHE

The total thermal resistance of MHPA–AWHE determines heat transfer performance. Here, the log mean temperature difference (LMTD) and the total thermal resistance (R) of indoor and outdoor MHPA–AWHEs are obtained according to the temperature distribution of the air and water sides. To analyze the balance and rationality of the combination of the indoor and outdoor MHPA–AWHEs of the SNCS, the ratio of the heat transfer units of the indoor to outdoor MHPA–AWHE is represented by N.

The LMTD of the indoor and outdoor MHPA–AWHEs is:

$$\mathrm{LMTD}=\frac{\left|T_{a,i}-T_{w,o}\right|-\left|T_{a,o}-T_{w.i}\right|}{ln\left(\left|T_{a,i}-T_{w,o}\right|/\left|T_{a,o}-T_{w.i}\right|\right)}.$$

(6)

The thermal resistance of indoor and outdoor MHPA–AWHEs is:

$$R_{IN}=\frac{{\mathrm{LMTD}}_{IN}}{Q_{IN}}$$

(7)

$$R_{OUT}=\frac{{\mathrm{LMTD}}_{OUT}}{Q_{OUT}}.$$

(8)

The ratio of the heat transfer units of indoor to outdoor MHPA–AWHE is:

$$N=\frac{n_{IN}}{n_{OUT}},$$

(9)

where n_IN and n_OUT are the rows of heat transfer units of indoor and outdoor MHPA–AWHEs, respectively.

2.6.3. Convection Heat Transfer of Air and Water Sides

The convective heat transfer coefficient of the air and water sides of the heat exchanger is influenced by different kinds of fins and structures. The greater the disturbance degree and velocity of the fluid are, the stronger the convective heat transfer ability is.

The convective heat transfer coefficient of the air-side is:

$$h_a=\frac{Q_a}{A_a(T_a-T_e)}.$$

(10)

The convective heat transfer coefficient of the water side is:

$$h_w=\frac{Q_w}{A_w(T_c-T_w)}.$$

(11)

2.6.4. EER of the SNCS

EER shows the refrigeration performance coefficient of the cooling system. The higher the EER is, the more heat is absorbed or the less electric power is consumed. Here, the electric power of the fan and pump is used as the input and the heat dissipation is the output of the SNCS. EER is used to evaluate the output–input ratio, as follows:

$$\mathrm{EER}=\frac{Q_{AVE}}{E_{IN,f}+E_{OUT,f}+E_p},$$

(12)

where E_IN,f and E_OUT,f correspond to the power consumption of fans of the SNCS, and E_p is the power consumption of the pump.

2.7. Uncertainty Analysis

The error caused by the precision of instruments and equipment in the experiment is called systematic error. The indirect error is calculated by error transfer formula. If y is a function of the independent variables x₁, x₂, x₃, …, x_n, the uncertainty of y is calculated as follows, and the results are shown in

Table 3. Uncertainty of the main parameter variables.

Physical Quantity	Numerical Range	Relative Uncertainty
QIN,a	1582–11,414 W	±3.16–6.61%
QIN,w	1694–12,639 W	±3.02–6.14%
QOUT,a	1551–11,357 W	±4.22–8.35%
QOUT,w	1635–12,248 W	±3.84–7.19%
Δβ	3.6–13.1%	±2.17–6.11%
RIN	1.72 × 10−3–2.05 × 10−3 K/W	±2.87–6.36%
ROUT	2.08 × 10−3–3.19 × 10−3 K/W	±4.87–6.13%
EER	2.88–12.7	±3.01–5.71%

:

$$\delta y={\left[{\left(\frac{\partial y}{\partial x_1}\delta x_1\right)}^2+{\left(\frac{\partial y}{\partial x_2}\delta x_2\right)}^2+\cdots +{\left(\frac{\partial y}{\partial x_n}\delta x_n\right)}^2\right]}^{\frac{1}{2}}.$$

(13)

3. Result and Discussion

3.1. Performance of the SNCS under Room-Based Condition

3.1.1. Heat Loss Performance

The heat losses under different inlet water temperatures are analyzed by changing the air and water flow rates. As shown in Figure 7, the heat transfer rate decreases linearly with the increase of the outdoor environment temperature. The average heat loss percentage is 8.7% when q_V,a is 3000 m³/h and q_V,w is 1200 L/h. As shown in

Table 4. Heat loss percentage Δβ under different conditions.

Condition	TOUT,a,i (°C)
	−15	−10	−5	0	5	10	15	AVE
1200 L/h, 3000 m3/h	7.3%	9.3%	8.2%	8.1%	7.8%	9.9%	10.7%	8.76%
1200 L/h, 2000 m3/h	8.1%	7.2%	7.1%	8.0%	6.1%	7.4%	7.3%	7.31%
1200 L/h, 1000 m3/h	6.8%	6.5%	7.1%	5.2%	5.6%	6.1%	6.2%	6.21%
800 L/h, 2000 m3/h	5.9%	6.7%	7.5%	8.1%	8.0%	7.6%	7.5%	7.33%

, heat loss percentages are influenced by q_V,a and have a slight influence on q_V,w. The larger the q_V,a is, the larger the heat loss percentage is. The cold air permeability and heat loss increase gradually with the increase of q_V,a. During the experiment, the heat loss percentage Δβ is always at a lower level, which is less than 10%.

3.1.2. Thermal Performance of SNCS at N = 1

The heat transfer performance of the SNCS at N = 1 is studied to analyze the influence of the flow rate more intuitively. Figure 8 and Figure 9 show the variation of the heat transfer rate and convection heat transfer coefficient under different flow rates and outdoor environment temperatures.

As shown in Figure 8, the outdoor environment temperature at −15 °C is selected for analysis. When q_V,a increases, the same trends of the heat transfer rate are observed at different outdoor temperatures. When q_V,a is less than 1000 m³/h, the heat transfer rate is at a relatively low level, and the convective heat transfer coefficient is below 46 W/(m²·K). As q_V,a increases to 1500 m³/h, the heat transfer rate increases by 34.5% more than that of 1000 m³/h. A larger slope (Slope 1) is noted at this stage; the disturbance of the air flow is enhanced by the serrated fins, and the convective heat transfer coefficient is 54.8 W/(m²·K). As q_V,a increases to 3000 m³/h, the growth of the convective heat transfer coefficient slows down, as shown in Slope 2.

As shown in Figure 9, the heat transfer rate increases by 11.5% from 6.68 to 7.45 kW when q_V,w increases from 400 to 1200 L/h. The convective heat transfer coefficient fluctuates between 469.7 and 518.2 W/(m²·K) when the outdoor temperature is −15 °C. Hence, q_V,w has little influence on heat transfer. The water side has typical laminar flow as the average Re after the calculation is 201 and is lower than the critical Reynolds number (Re = 2300) in the tube during this experiment.

The results show that q_V,a is the main factor affecting the heat transfer rate of the SNCS. However, with the increase of outdoor temperatures, the heat transfer performance of the heat exchanger evidently decreases.

According to the low heat transfer rate at higher environment temperatures, the temperature and thermal resistance distribution of SNCS at N = 1 during the heat transfer process are analyzed and optimized. The curve represents the heat transfer process and the arrow represents the direction of heat transfer. The difference between the y- and the x-coordinates of any two points on the curve is the temperature difference and the heat transfer in this process, respectively, as shown in Figure 10. The conditions of q_V,w = 1200 L/h and q_V,a = 2500 m³/h are selected.

When the outdoor inlet air temperature is −15 °C, the average temperatures of the air-side of the indoor and outdoor MHPA–AWHEs are 23.86 °C and −10.95 °C. The average temperatures of the water-side of the indoor and outdoor MHPA–AWHEs are 11.17 °C and 11.13 °C, respectively, as shown in Figure 10a. The small heat loss of the water sides from the indoor to outdoor MHPA–AWHE is observed, as the temperature difference is only 0.04 °C. However, the temperature of the cooling water is closer to the air temperature of the indoor side than that of the outdoor side. Moreover, the LMTDs of the indoor and outdoor MHPA–AWHEs are 12.69 °C and 22.08 °C, respectively. The total heat transfer thermal resistance of the outdoor MHPA–AWHE is 1.76 times larger than that of the indoor MHPA–AWHE.

As shown in Figure 10b, when the outdoor inlet air temperature is 5 °C, the total heat transfer rate is 4.19 kW, and the temperature of the cooling water is closer to the air temperature of the indoor side than that of outdoor side. Both the LMTD and the total heat transfer thermal resistance of the outdoor MHPA–AWHE are 1.64 times larger than that of the indoor MHPA–AWHE.

Therefore, the heat transfer rate of the outdoor MHPA–AWHE is insufficient, and the combination form of the SNCS is defective. Moreover, the greater the temperature difference between indoor and outdoor is, the more evident the imbalance is. Increasing the rows of the heat transfer units of the outdoor MHPA–AWHE is necessary to enhance the performance of the SNCS.

3.2. Optimization Analysis of SNCS at Different N

3.2.1. Heat Transfer Performance of SNCS at Different N

The form of the indoor MHPA–AWHE composed of 12 rows of heat transfer units remained unchanged. The heat transfer performances of SNCS at N = 1, N = 0.75, and N = 0.6 were compared.

As shown in Figure 11, the heat transfer rate of SNCS at different N evidently increased with the increase of the q_V,a. When N = 1, the heat transfer rate finally tended to be gentle. When N = 0.75 and N = 0.6, the heat transfer rates still had a large growth trend during the experiment. As shown in Figure 11a, for example, when the outdoor inlet air temperature was −15 °C, the maximum heat transfer rates of N = 1, N = 0.75, and N = 0.6 were 7.63, 8.70, and 8.91 kW, respectively. The average heat transfer rate under different q_V,a of N = 0.75 and N = 0.6 were 13.1% and 14.3% higher than that of N = 1, respectively. The performance under other outdoor inlet air temperatures is seen in

Table 5. Heat transfer rate under different forms.

Different Forms of SNCS	TOUT,a,i = −15 °C		TOUT,a,i = −5 °C		TOUT,a,i = 10 °C
	Qmax (kW)	Increase Rate	Qmax (kW)	Increase Rate	Qmax (kW)	Increase Rate
N = 1	7.63	-	5.92	-	3.29	-
N = 0.75	8.70	17.8%	7.02	17.8%	4.39	29.6%
N = 0.6	8.91	18.1%	7.04	18.1%	4.37	29.2%

. The smaller the indoor and outdoor temperature difference was, the more evident the difference of the heat transfer rate was.

As shown in Figure 11d, the increase of q_V,w occurs when the outdoor inlet air temperatures were −15 °C, −5 °C, and 10 °C. The average heat transfer rates under different water flow rates of N = 0.75 are 10.8%, 14.7%, and 25.2% larger than that of N = 1, respectively. Moreover, the average heat transfer rates under different q_V,w of N = 0.6 are nearly the same. Hence, the heat transfer rate basically reached its maximum value when N = 0.75. Thus, continuing to increase the rows of the core component of the outdoor MHPA–AWHE is not helpful.

As can be seen from the temperature distribution diagram in Figure 12a, q_V,a = 2500 m³/h and q_V,w = 1200 L/h are selected. When the N of the SNCS was 0.75, the LMTD of the indoor and outdoor MHPA–AWHEs were 16.34 °C and 17.07 °C, respectively, and the total heat transfer resistances were 2.02 × 10⁻³ and 2.11 × 10⁻³ K/W, respectively. Both indoor and outdoor MHPA–AWHEs were at the same level, which indicated that the heat transfer process of each part of the SNCS reached a balance. When the N of the SNCS is at 0.6, as shown in Figure 12b, little difference is observed in the heat transfer rate and the temperature distribution between N = 0.75 and N = 0.6.

According to the above optimization analysis of SNCS at different N, the combination form of SNCS at N = 0.75 was more reasonable and had a better heat transfer performance with the balanced temperature and thermal resistance distribution.

3.2.2. Flow Characteristics of the Indoor and Outdoor MHPA–AWHEs

The pressure drop directly determined the energy consumption of fans and pumps. The pressure drop and power consumption characteristics of SNCS at different N were analyzed. As shown in Figure 13, the fitting curve of the pressure drop of the air-side increased exponentially, and the power consumption of the fans increased linearly, with q_V,a increasing from 1000 to 3000 m³/h. The fitting curves are shown in

Table 6. Fitting curves of indoor and outdoor MHPA–AWHEs at different N.

Curve	Condition	Fitting Curve of Pressure Drop	R2
Curve 1	Indoor, 12 rows	$\Delta P_a=0.0000243q_{V,a}^2+0.0401q_{V,a}$	0.99984
Curve 2	Outdoor, 12 rows	$\Delta P_a=0.0000247q_{V,a}^2+0.042q_{V,a}$	0.99992
Curve 3	Outdoor, 16 rows	$\Delta P_a=0.0000182q_{V,a}^2+0.0378q_{V,a}$	0.99949
Curve 4	Outdoor, 20 rows	$\Delta P_a=0.0000108q_{V,a}^2+0.035q_{V,a}$	0.99884

. The pressure drop and power consumption of the indoor and outdoor MHPA–AWHEs with 12 rows were basically equal due to the same windward area and air velocity. With the increase of the rows of the outdoor MHPA–AWHE, which caused the windward area to increase and velocity to decrease, the pressure drop and power consumption decrease correspondingly. When q_V,a was 3000 m³/h, the maximum pressure drops of outdoor MHPA–AWHE with 12, 16, and 20 rows were 352, 276, and 198 Pa, respectively. The maximum fan power consumption of outdoor MHPA–AWHE with 12, 16, and 20 rows was 296, 268, and 233 W, respectively. The pressure drop and power consumption of MHPA–AWHE with 16 rows were reduced by 27.5% and 8.5% compared with that of MHPA–AWHE with 12 rows.

Figure 14 shows that the fitting curve of the pressure drop of the water-side increases exponentially, and the power consumption of the pump increased linearly with q_V,w increasing from 400 to 1200 L/h. The fitting curves of the water side of the SNCS at different N were calculated in

Table 7. Fitting curves of pressure drop of water side at different N.

Curves	Condition	Fitting Equation of Pressure Drop	R2
Curve 5	N = 1	$\Delta P_w=1.32e^{q_{V,w}/864.7}+10.88$	0.99725
Curve 6	N = 0.75	$\Delta P_w=1.76e^{q_{V,w}/782.1}+17.90$	0.99666
Curve 7	N = 0.6	$\Delta P_w=2.79e^{q_{V,w}/858.5}+23.26$	0.99949

. In contrast, with the increase of the rows of the outdoor heat transfer units, the pressure drop and power consumption increased correspondingly. During the experiment, when q_V,w = 1200 L/h, the maximum pressure drops of SNCS at N = 1, N = 0.75, and N = 0.6 were 16.2, 26.1, and 34.6 kPa, respectively. As the rows increase, the number of distribution pipes increased with the local pressure drop. The maximum power consumption of water pumps with 12, 16, and 20 rows was 189, 243, and 288 W, respectively. The pressure drop and power consumption with 16 rows were increased by 61.1% and 28.6%, respectively, compared with that of MHPA–AWHE with 12 rows.

In summary, the pressure drop and power consumption with different air and water flow rates were within the reasonable range of the experiment. With the increase of the rows of heat transfer units, the influence of the pressure drop and power consumption on the water-side was more evident than that on the air-side. The fitting curve of the pressure drop had theoretical support in the heat exchanger design and practical engineering application.

3.2.3. Energy Efficiency Ratio (EER)

The comprehensive performance of SNCS was evaluated by the EER of the heat transfer rate with the pressure drop and total power consumption of fans and pumps.

As shown in Figure 15a–c, q_V,w was maintained at 1200 L/h, and the EER of SNCS with different N had the same trend under different q_V,a. The EER increased slightly when q_V,a was less than 1500 m³/h. As the destruction by serrated fins to the air boundary layer was intensified, the heat transfer was enhanced. The increasing heat transfer rate was greater than the increasing power consumption. As q_V,a continued to increase from 1500 to 3000 m³/h, the EER decreased linearly. The increasing pressure drop dominated the comprehensive performance, and the increasing power consumption was larger than the increasing heat transfer rate. When the SNCS at N = 1, N = 0.75, and N = 0.6, the maximum EER was 13.44, 14.01, and 13.2, respectively, when the outdoor temperature is −15 °C. With the increase of rows of heat transfer units, the power consumption of the fans decreased, whereas the power consumption of the water pump increased. The SNCS at N = 0.75 had the largest EER during the experiment.

Figure 15d shows that with the increase of q_V,w, EER of the SNCS at different N decreased slightly. Hence, the water flow rate had little influence on the EER of the SNCS at the same N and the same environment temperature. When the outdoor temperature was −15 °C, the average EERs of the SNCS at N = 1, N = 0.75, and N = 0.6 were 11.09, 11.89, and 11.7, respectively. The EER of the SNCS at N = 0.75 was always at the highest level under different environment temperatures.

In the above series of studies, we indicated that under different conditions, EER_N=1 < EER_N=0.6 < EER_N=0.75, the SNCS at N = 0.75 had the best comprehensive performance under different air and water flow rates and outdoor environment temperatures.

3.3. Thermal Performance of SNCS under Heat Channel-Based Condition

Based on the best performance of the SNCS at N = 0.75, this paper proposed the heat dissipation condition at 38 °C in a closed heat channel between server cabinets. The working time of the air conditioner of the SNCS could be further reduced by increasing the range of available temperature difference between indoor and outdoor sides to enhance heat transfer capacity. Given that q_V,w had little influence on heat transfer performance, and the pressure drop was basically independent of temperature, the current study focused on the heat transfer performance of two heat dissipations under different q_V,a.

As shown in Figure 16, the heat transfer rate of heat dissipation at 38 °C significantly improves compared with that at 28 °C. q_V,w was kept at 1200 L/h when q_V,a were 1000, 2000, and 3000 m³/h. The maximum heat transfer rates of heat dissipation at 38 °C were 6.72, 9.96, and 12.4 kW, which increased by 14.3%, 32.6%, and 42.5% compared to the heat dissipation at 28 °C, respectively. The larger the temperature difference between indoor and outdoor was, the larger the increase of heat transfer rate of heat dissipation at 38 °C was compared with that at 28 °C. MHPA had a larger thermal conductivity performance under the higher evaporation temperature.

The comprehensive performances were compared under different outdoor temperatures and two heat dissipation conditions. As shown in Figure 17, with the increase of the outdoor temperature, the variation of the EER under two conditions decreased linearly. The overall level of the EER at 38 °C was always higher than that at 28 °C. When q_V,a was 1500 m³/h and the outdoor temperature was −15 °C, the maximum EER at 38 °C was 17.15 due to the higher available temperature difference between heat resource and outdoor temperature. This value was an increase of 22.64% compared with the heat dissipation at 28 °C. The fitting curve of the system EER under different environment temperatures of a standard air conditioner (AC) in a data center was obtained through Equation (14). The average EER of a standard AC was 4.89 when the outdoor temperature varied from −15 °C to 15 °C. When the outdoor temperature was higher than 15 °C, the EER of the room-based condition was lower than AC, and the SNCS was no longer recommended. However, when the EER of the heat channel-based condition was always higher than AC, the SNCS still showed a better performance and could be used in a wider temperature range.

$$\mathrm{EER}=5.273-0.01125T_{OUT}-0.00049T_{OUT}^2-0.000012T_{OUT}^3$$

(14)

During the experiment, when q_V,a were 1500, 2000, and 2500 m³/h, the SNCS had a larger EER. When q_V,a was 1000 m³/h with a lower heat transfer rate and 3000 m³/h with a lower EER, the SNCS was not recommended. The fitting curves of the EER of the SNCS module under two conditions were obtained, as shown in

Table 8. Fitting curves of EER of the SNCS module under two conditions.

Heat Dissipation Conditions	qV,a (m3/h)	qV,w (L/h)	Fitting Curves of EER	R2
Room-based	1500	1200	$EER=12.763-0.3033T_{OUT}$	0.9965
	2000	1200	$EER=12.242-0.3046T_{OUT}$	0.9975
	2500	1200	$EER=11.731-0.3011T_{OUT}$	0.9981
Heat channel-based	1500	1200	$EER=9.739-0.288T_{OUT}$	0.9975
	2000	1200	$EER=9.011-0.249T_{OUT}$	0.9969
	2500	1200	$EER=8.199-0.228T_{OUT}$	0.9981

. The following data provide a reference for the practical application of the curves in different regions.

4. Conclusions

A new kind of split natural cooling system (SNCS) based on a micro heat pipe array that uses natural cold energy efficiently was proposed. The performance of the SNCS was analyzed and optimized comprehensively based on the heat transfer and flow characteristics.

(1) During the heat transfer process, heat loss was less and the air flow rate had a greater influence on the heat transfer performance of the SNCS than the water flow rate. When the air flow rate was larger than 1500 m³/h, the heat transfer was enhanced due to the destruction of serrated fins to the air boundary layer;

(2) The ratio of heat transfer components (N) at about N = 0.75 showed the best thermal performance. The temperatures and thermal resistance distributions of indoor and outdoor MHPA–AWHE were reasonably balanced. The average heat transfer rates were 13.1% higher than that of N = 1;

(3) The pressure drop of the SNCS at N = 0.75 was at a lower level. The maximum pressure drops of the air- and water-sides were 276 Pa and 26.1 kPa, respectively. The fitting curve of the pressure drop was obtained to provide a reference for the heat exchanger design and equipment selection;

(4) Two kinds of heat dissipation conditions—room and heat channel-based—were proposed. The maximum EERs of the heat dissipation at 38 °C and 28 °C were 17.15 and 13.98, respectively, when the air flow rate was 1500 m³/h and the outdoor temperature was −15 °C. The SNCS under the room-based condition was no longer applicable when the outdoor temperature was higher than 15 °C. The SNCS under the heat channel-based condition could be used in a wider temperature range. The fitting curves of the EER of the SNCS module were obtained as a reference for its practical application in different regions.