Professional Documents
Culture Documents
Energy
journal homepage: www.elsevier.com/locate/energy
a r t i c l e i n f o a b s t r a c t
Article history: The demand for the optimization of cooling water systems to reduce energy consumption is increasing
Received 8 March 2021 nowadays. Based on the heat transfer performance of a cooling tower, this paper proposed a heat transfer
Received in revised form model for cooling towers. To verify the accuracy of the model and to demonstrate the applicability of off-
28 April 2021
design conditions, an experimental system for evaluating the air/water heat transfer performance of
Accepted 6 May 2021
cooling towers was established. The experimental results showed that the proposed model for cooling
Available online 12 May 2021
towers had an ability of high-accuracy prediction and adaptability to off-design conditions. Based on the
proposed model for cooling towers, the mathematical model of cooling water system online optimization
Keywords:
Cooling tower
is established. The hybrid programming particle swarm optimization (HPePSO) algorithm to optimize
Cooling water system the operation of cooling water systems was proposed to solve the online optimization problem of
HP-PSO algorithm continuous variables and discrete variables in cooling water systems. The advantages of the HP-PSO
Online optimization method algorithm were verified by the simulation of a small data center. The simulation results showed that
Energy-saving the proposed online optimization method of cooling water systems can reduce energy consumption by
15.3% comparing with other method. This paper can provide an important reference for the operation
optimization of cooling water systems in practical engineering.
© 2021 Elsevier Ltd. All rights reserved.
https://doi.org/10.1016/j.energy.2021.120896
0360-5442/© 2021 Elsevier Ltd. All rights reserved.
K. Ma, M. Liu and J. Zhang Energy 231 (2021) 120896
equation, which simplified the performance prediction of cooling Using the temperature optimization method to adjust the speed
towers. In the 1970s, Rogener proposed the Popper model by of cooling water pumps and cooling tower fans can reduce the
setting the Lewis number as a variable to accurately calculate the air energy consumption of cooling water systems to a certain extent.
state at the outlet of cooling towers. But the calculation was rela- However, the reduction in energy consumption of cooling water
tively complex and difficult to apply [9]. Since the 21st century, systems is based on the temperature constraint which cannot
some researchers have established some simplified heat transfer ensure the optimal operating state of cooling water systems. The
models to predict the heat transfer performance of cooling towers. application of an optimization intelligent algorithm can conve-
Jin et al. [10] established a simplified calculation model for cooling niently optimize the operation of cooling water systems.
towers based on the number of ε-NTU model through the energy Chow et al. [19] proposed the method of combining a neural
balance and the heat and mass transfer analysis, and the experi- network with a genetic algorithm (GA) to solve the global optimi-
mental data verified that the model can predict the performance of zation of absorption chillers. Lu et al. [20] used an improved GA to
cooling towers simply and accurately. Asvpoositkul [11] proposed a optimize the operation of a cooling water system for central air
simple model to describe the performance of cooling towers by the conditioning, which greatly reduced the operation cost of this
mass evaporation rate equation. When the flow rate and the inlet cooling water system. Ma et al. [21] used a GA as a global optimi-
water temperature were closer to the design conditions, the accu- zation tool to search for the optimal control settings of model-
racy of the model results was higher. Bornman et al. [12] proposed a based central water chillers to save the daily energy consumption
model based on the principle of energy balance to predict heat of the system. Seo et al. [22] determined the optimal operation
transfer without iterative calculations. The results showed that the mode of HVAC (Heating, Ventilation, and Air Conditioning) systems
average prediction error of the outlet water temperature was be- by using the improved multi-island GA to minimize energy con-
tween 0.5 C and 0.7 C. sumption. Chang [23] and Ardakani et al. [24] used GA and the
The Merkel model, the ε-NTU model, and other models have particle swarm optimization respectively to solve the problem of
good applicability in the design and calculation, but iteration is the optimization of chiller operation. Kusiak et al. [25] established
needed to obtain the heat transfer model for cooling towers. The the nonlinear relationship among energy consumption, the control
calculation processes of these models are complex, and it is difficult parameters, and the state parameters in HVAC systems and used
to apply to the online operation optimization of cooling water the PSO algorithm to optimize the system. Beghi et al. [26] pro-
systems. Although other simplified models have simple calculation posed an efficient multi-unit management method based on the
structures, they are only suitable for limited working conditions. particle swarm optimization for HVAC systems with medium and
When the operation of cooling towers deviates from the design high cooling capacities, which effectively solved this kind of
conditions or the model regression conditions, it is difficult for the nonlinear constrained optimization problem. Lee et al. [27] used
performance model for cooling towers to be applied to the online the hybrid optimization algorithm of the particle swarm optimi-
optimization control. Therefore, the cooling tower model suitable zation and the Hooke Jeeves algorithm to determine the optimal
for online operation optimization is still lacking, which needs to be temperature setting for chilled water systems. Mu et al. [28] pro-
studied further. posed the multivariable Newton base optimization (ESC) method,
The operation optimization of cooling water systems refers to which took the total power of a cooling water system as the feed-
the reasonable adjustment of the regulating parameters in cooling back signal to find the optimal output of the fan speed in a cooling
water systems to reduce energy consumption. The temperature tower and the flow rate of the cooling water to achieve the goal of
difference method is one of the most popular methods for opti- the energy-saving. Singh et al. [29] proposed a feedback model
mizing the operation of cooling water systems. based on the constrained inverse parameter inversion, which took
Nasrabadi et al. [13] studied two strategies, an increase in the the mass flow rates of air and water as control variables to control
airflow rate of cooling towers and a reduction in the water flow the output performance of a cooling tower, so that the system could
rate. The results showed that in a warm and humid temperate meet the heat removal rate of the cooling tower and ensure the
climate, the assumption of 2K temperature difference was more minimum power consumption. Keshtkar et al. [30] adopted the
suitable for a wider range of weather conditions than 1K. Huang MCDM-TOPSIS method based on the Popper thermal calculation
et al. [14] adopted the predictive control method for the approach model, and the results showed that the method could obtain the
temperature to optimize the set value of the condensate temper- minimum energy consumption of air and water flows and the op-
ature, which saved about 9.67% of energy consumption in chillers timum working conditions for the operation of cooling towers.
and cooling towers. F. W. Yu et al. [15] studied the system optimi- As the air conditioning refrigeration system operates under
zation method for water-cooled chillers based on the speed regu- variable conditions, the optimal value of the regulating variable will
lation of the cooling load. By adjusting the speed of fans and pumps change according to the working condition and generally shows a
through cooling loads, the COP of the system could increase by nonlinear relationship. These intelligent optimization methods are
1.4e16.1% compared with the equivalent system with a fixed tem- more suitable for the operation optimization of the system under
perature and flow rate. Cortinovis et al. [16] established the oper- variable conditions so that the system can achieve maximum
ation optimization model for cooling water systems including energy-saving. However, the intelligent optimization algorithms
cooling towers, pipeline networks, and heat exchangers. When a used in the existing researches have not been combined with the
low temperature was not required, the set value of the outlet characteristics of the optimization model for cooling water sys-
temperature in cooling towers was higher. When a low tempera- tems. Therefore, most of the optimization methods are blind, the
ture was required, the flow rates of the cooling return water and air calculation processes are complex and time-consuming, and the
in cooling towers increased. Marques et al. [17] proposed the split algorithms are not suitable for online applications. Also, the syn-
range control strategy for the water temperature which can chronous optimization of discrete variables and continuous vari-
improve the efficiency of open cooling towers. Chang et al. [18] ables has been less considered in the existent algorithms for the
tried to reduce the number of fan start-up and stop times for optimal operation of cooling water systems and there is a lack of
reducing energy consumption. They applied the temperature zone the optimized methods for the operating units of equipment in the
method instead of the fan set value to control the outlet water cooling water system.
temperature of a cooling tower at 32 C and the online test To solve the above problems, this paper proposed a heat transfer
confirmed the energy-saving effect of this method. model of cooling towers which is suit for the online optimization,
2
K. Ma, M. Liu and J. Zhang Energy 231 (2021) 120896
and the accuracy and the applicability of this heat transfer model Where kxv is the volumetric mass transfer coefficient (kg/m3/s), tw
for cooling towers was verified by experiments. Based on the heat is the water temperature ( C), A is the cross area of the packing
transfer model for cooling towers, the mathematical model to (m2), Z is the heigh of the packing (m), Dhm is the average enthalpy
optimize the operation of cooling water systems online was difference in the heat transfer process (kJ/kg), l is the air-water
established. Also, the HP-PSO algorithm was proposed to solve the mass ratio. Combining Eq. (1) with Eq. (5), Eq. (6) is obtained:
optimization problems of continuous variables and discrete vari-
pmpw1
ables in the operation of cooling water systems. Through the sys-
Qc ¼ p1 h}wi hai ¼ r , mm
w ,ma Dhm AZ
n
(6)
tem simulation method, the optimization of the algorithm was mw
verified, and the optimization results were analyzed in regards to 1 þ p2 ma
energy-saving. The optimization algorithm proposed in this paper
can provide a feasible reference for the practical engineering Thus Dhm can be simplified as follows:
application of cooling water systems. p
pmw1
p m
p mw1 mm1
AZrmm
w ma
1m
(7)
2.1. Heat transfer model for cooling towers
According to the Merkel enthalpy difference equation, the outlet
The establishment of the heat transfer model of cooling tower is water temperature of the cooling tower is only related to the air/
based on the following assumptions: water ratio in the cooling tower. Based on assumption 5), Dhm can
be shown as follows:
1) The heat and mass transfer processes in cooling towers are only
ð twi
along the direction of air and water flows;
h}w ha h}wi hao þ h}wo hai
2) The heat and mass transfer processes are steady; two
Dhm ¼ ¼ (8)
3) The heat and mass transfer processes from cooling towers to the Dtw 2
surrounding environment are ignored;
4) The sensible heat lost by air is equal to the latent heat required cw ðtwi two Þ
for water evaporation; hao ¼ hai þ (9)
Kl
5) The average enthalpy difference of the cooling process is only
related to the temperature difference between the inlet and the
h}wo ¼ k,two þ b (10)
outlet water temperatures and the air/water ratio.
Where the subscripts i and o represent the inlet and the outlet of
According to the analysis of the internal convective heat transfer
the cooling towers, respectively. According to Eqs. (8)e(10), the
principle in the cooling tower, the heat transfer process in the
average enthalpy difference in the cooling tower Dhm has no rela-
cooling tower can be expressed as follows [10]:
tionship with a single air volume or water volume, but is only
pmpw1 related to the air/water ratio and the inlet enthalpy. By combining
Qc ¼ p1 h}wi hai (1) Eq. (7) and Eq. (8), the dimensional equilibrium relation about mw is
1 þ p2 mma
w obtained, as shown in Eq. (11):
p1 m ¼ 1 m (11)
Where Qc is the heat transfer capacity of cooling towers (kW), ma
and mw are the mass flow rate of air and water (kg/s), h}wi is the Therefore, the value of p1 is equal to 1. After the above deriva-
tion, the heat transfer capacity of the cooling tower can be calcu-
saturated enthalpy of wet air corresponding to the inlet water
lated as follows:
temperature of the cooling towers (kJ/kg), hai is the enthalpy of the
inlet air of cooling towers (kJ/kg). p0, p1, and p2 are the parameters p0 mw
to be regressed in Eq. (1), which are related to cooling tower Qc ¼ ðtwi twb Þ (12)
specification, the inlet water temperature and the outdoor wet- 1 þ p2 m w
ma
bulb temperature of the cooling tower. Therefore, this model
needs to be improved. Where twb is the wet-bulb temperature of the air ( C). Although the
According to the Merkel enthalpy difference model, the rela- regression values of p0 and p2 change with the change in the out-
tionship between the Me number and the enthalpy is as follows: door environment temperature and the inlet water temperature of
tð
the cooling tower, the difference between them is always very
wi
kxv AZ cw dtw small. Therefore, assuming that the values of p0 and p2 are equal,
Me ¼ ¼ 00 (2) the heat transfer capacity of the cooling tower can be simplified as
mw K hw ha
two follows:
p0 mw
kxv ¼ r,mm n
w ,ma (3) Qc ¼ ðtwi twb Þ (13)
1 þ p0 m w
ma
kxv AZ ma n
Me ¼ ¼ C,mm1 n
w ,ma ¼ C ¼ C ln (4) Among the regression values, p0 is a parameter related to the
mw mw
outdoor wet-bulb temperature and the inlet water temperature of
the cooling tower. Referring to the correction mode of the volu-
Qc ¼ kxv Dhm AZ ¼ r,mm
w ,ma Dhm AZ
n
(5)
metric mass transfer coefficient to the inlet water temperature, p0
can be expressed as follows:
3
K. Ma, M. Liu and J. Zhang Energy 231 (2021) 120896
The GN model was used to predict the power consumption of a The schematic diagram of the experimental system for evalu-
chiller [31]. The COP of the chiller can be expressed as Eq. (18). The ating the heat transfer performance of a cooling tower is shown in
relationship between the input power and the cooling capacity of Fig. 1.
the chiller can be expressed as Eq. (19). The experimental system mainly includes the following parts:
1 tci 1 qc tci (1) Packing and other accessories
¼ 1þ þ qe þ f (18)
COP teo Qe teo
The S-shaped corrugated plastic film which can increase the
t heat transfer and reduce the airflow resistance was selected as the
Pin ¼ ðQe þ qe Þ ci þ ðf 1ÞQe qc (19) packing of the cooling tower. In Fig. 2(a), the size of each packing
teo
piece was 300 400 400 mm, and 3 pieces were used totally. The
Where Qe is the heat transfer capacity of the evaporator (kW), qe is height of the test packing was 0.9 m and the cross-sectional area of
the heat loss of the evaporator (kW), qc is the heat loss of the the packing body was 0.16 m2. To spray the cooling water evenly on
condenser (kW), f is the correction parameter of the pipeline the surface of the packing, the conical nozzles were selected and
fouling, tci is the inlet water temperature of the condenser ( C), teo is they were evenly arranged above the packing as shown in Fig. 2(b).
the outlet water temperature of the evaporator ( C). A wave-shaped drift eliminator was installed above each nozzle to
remove the droplets in the air as shown in Fig. 2(c).
2.3. Power model for cooling water pump and fan
(2) Water pumps, valves and collecting tank
The head of the cooling water pump was used to provide a
certain water flow rate and overcome the height difference. To realize the continuous regulation of the cooling water flow in
Therefore, the head of the cooling water pump can be expressed as the heat transfer process, the water pump installed on the inlet
follows: pipe of the cooling tower was a frequency-conversion water pump.
The water pump installed on the outlet pipe of the cooling tower
H ¼ SL2 þ DH (20) was a constant-frequency water pump as shown in Fig. 3(a). The
outlet water pipe of the cooling tower was equipped with an
In Eq. (20), H is the pump head (m), S is the pipe network electric regulating valve to ensure the cooling water circulation as
impedance of the cooling water system (s2/m5), L is the water shown in Fig. 3(b). A collecting tank of 800 800 mm square was
volume flow rate (m3/s), DH is the difference in height between the set under the packing to collect the cooling water after spraying as
cooling water outlet and the liquid level of the collecting tank (m). shown in Fig. 3(c). A water level gauge was set outside the col-
The shaft power Pe of the cooling water pump can be calculated lecting tank to control the water level.
according to Eq. (21).
(3) Axial flow fan and duct details
Pe ¼ rgLH (21)
The actual power consumption P of the pump is expressed by The flow of the air system adopted the way of upper air
4
K. Ma, M. Liu and J. Zhang Energy 231 (2021) 120896
Fig. 1. Schematic diagram of the experimental system for evaluating the heat transfer performance of a cooling tower.
induction. A PPR pipe with DN300 was used as the air duct as meter. The flow rate of the cooling water was measured by an
shown in Fig. 4(a). The axial flow fan was selected to supply the electromagnetic flowmeter and the airflow rate was measured by
airflow power at the end of the air duct as shown in Fig. 4(b). an orifice flowmeter with a differential pressure transmitter.
The measurement range and accuracy of all the measuring de-
(4) Thermostatic water tank and electric heaters vices in the test system are shown in Table 1.
A constant-temperature water tank was used as the heat source. 3.2. Energy balance of the experiment system
The PID temperature controller was used to automatically adjust
the power output of the electric heaters to ensure the constant To ensure the reliability of the test results, it was necessary to
water temperature at the outlet of the water tank. The water tank verify the energy balance error of the cooling tower test platform
and the electric heaters are shown in Fig. 5(a) and the PID tem- after construction. According to the energy conservation equation,
perature controller is shown in Fig. 5(b). the heat lost from the cooling water is equal to the heat gained by
air in the cooling tower. However, due to the heat transfer between
(5) Measuring equipment the wall and air, the heat transfer between the wall and the cooling
water, the solar radiation, and the flow loss, the heat loss from the
The temperature of the cooling water was measured by a ther- cooling water was different from the heat gained by air. The heat
mal resistance temperature measuring device (PT-100) and the air loss Qw of the cooling water and the heat gained by air Qa in the
temperature was measured by a thermocouple temperature heat transfer process of the cooling tower is as follows:
measuring device with a fast response speed. The temperature
measuring devices were calibrated by the standard constant tem- 1
perature calibrator using ethylene glycol. The humidity of air was Qw ¼ cw mw ðtwi two Þ (26)
K
measured by a humidity transmitter calibrated by a dew point
5
K. Ma, M. Liu and J. Zhang Energy 231 (2021) 120896
Fig. 4. Axial flow fan and duct details: (a) duct; (b) axial flow fan.
Fig. 5. Details of the thermostatic water tank and the electric heaters.
Qa ¼ ma ðhao hai Þ (27)
The energy balance error DQe can be expressed as: 4. Optimization model and method
6
K. Ma, M. Liu and J. Zhang Energy 231 (2021) 120896
Table 1
Measurement parameters and accuracy of all the measuring devices used in the experimental system.
results in the operation optimization of HVAC systems and the variable optimization. The calculation rule of the discrete PSO al-
optimized design of cooling towers [33,34]. Therefore, considering gorithm is different from that of the continuous PSO algorithm. The
the applicability of the algorithm, the number of iterations, the calculation rules of the discrete PSO algorithm are as follows:
length of the calculation, and the difficulty of implementation, this
paper selected the PSO algorithm as the basic optimization (1) Subtraction of position
algorithm.
The basic theory of PSO is to initialize a group of particles in the The result of the subtraction between two positions is the ve-
D-dimensional space of a feasible solution. Each particle represents locity, that is, uki;d ¼ ykþ1 -yki;d . It means that in the k-th iteration, if
i;d
a potential optimal solution to the extremum optimization prob-
the velocity a is applied to position B, position C of the kþ1-th
lem. The position, velocity, and fitness were used to describe the
iteration will be obtained. In the k-th iteration, the velocity ob-
characteristics of particles. The particles combine their memory
tained by the position subtraction is calculated according to Eq.
information with the shared information within the population and
(43).
constantly change the velocity and the position. By comparing the
fitness of the individual extremum and the group extremum be- 8
tween the new particle and the original particle, the individual <0 ykþ1 ¼ yki;d
i;d
extremum and the group extremum are updated to complete the uki;d ¼ (43)
: ykþ1 ykþ1 syki;d
search for the optimal solution. i;d i;d
The update equations of the position and the velocity of the
particle i are given by Eq. (39) and Eq. (40), respectively. The
renewal equations of the individual extreme value and the total
extreme value are expressed by Eq. (41) and Eq. (42), respectively. (2) Addition of position and velocity
xkþ1
id
¼ xkid þ vkþ1
id
(39) The result of the addition operation of the particle position and
the velocity is to realize the particle position movement, that is
ykþ1 ¼ yki;d þuki;d . The addition of the position and the velocity is
vkþ1
id
¼ uvkid þ c1 r1 pkid xkid þ c2 r2 pkgd xkid (40) i;d
calculated according to Eq. (44).
8
< pkid f xkþ1 f xkid 8
id <∅ uki;d ¼ 0
pkþ1 ¼ (41) ykþ1 ¼
id : xk f xkþ1 < f xkid i;d : swap yki;d ; uki;d
(44)
id id uki;d s0
n o n o
pkþ1
gd
2 pk1d ; pk2d ; …; pkId ¼ min f pk1d ; f pk2d ; …; f pkId
(42) (3) Number multiplication of the velocity
Wherexkid is the position component of the k-th iteration of the The number multiplication of the particle velocity is expressed
particle i in the d dimension, vkid is the velocity component of the k- as ukþ1 ¼ c,uki;d . In the k-th iteration, the particle generates a
i;d
th iteration of the particle i in the d dimension, pkid is the component random number rand on the d-dimensional component of each
of the position of the historical optimal solution of the particle i in velocity. If rand is less than c, the original velocity is maintained. If
the d dimension, pkgd is the component of the position of the his- rand is greater than or equal to c, the d-dimensional velocity
component of the particle i is 0 in the kþ1-th iteration. The number
torical optimal solution currently found by all particles in the
multiplication of the velocity is calculated according to Eq. (45).
population in the d dimension, uvkid is the inertial part of the last
particle flight velocity, c1 r1 ðpkid xkid Þ is the individual cognitive part (
of a particle, that is, the distance between the current position of uki;d rand < c
ukþ1 ¼ (45)
i;d
0 rand c
the particle i and its optimal position, c2 r2 pkgd xkid is the social
cognitive part of a particle, that is, the distance between the current
position of the particle i and the optimal position of the group.
In Eq. (40), c1 and c2 are velocity accelerations which are used to
(4) Addition of the velocity
adjust the maximum learning step, r1 and r2 are random numbers
in the range of [0,1] to increase the search randomness, u is the
The result of the addition of the particle velocity is the new
inertia weight which is used to adjust the search range of the
velocity. The velocity addition is used to ensure the diversity of the
optimal solution space. In order to improve the optimization ability,
particle swarm. The addition of the velocity is calculated according
u is assigned as a linear decreasing number of 0.9e0.4, c1 is
to Eq. (46).
assigned as a linear decreasing number of 2.5e0.5 and c2 is
assigned a linear increasing number of 0.5e2.5. 8
>
>
>
> ua uai;d s0∩ubi;d ¼ 0 ∪
< i;d
4.2.2. Hybrid programming particle swarm optimization
ui;d ¼ uai;d s0∩ubi;d s0∩rand < 0:5 (46)
At present, the PSO algorithm is widely used in the optimization >
>
>
> b
of continuous functions while the researches on the optimization of : ui;d else
discrete functions are less. Combining the discrete PSO algorithm
and the continuous PSO algorithm, the optimization process can be The flow chart of the hybrid programming particle swarm
divided into two parts: combinatorial optimization and continuous optimization algorithm is shown in Fig. 7.
8
K. Ma, M. Liu and J. Zhang Energy 231 (2021) 120896
4.3.1. Simulation system details 5.1. Experimental regression results and verification of the cooling
To obtain the results of the operation optimization of cooling tower model
water systems, a typical cooling water system for air conditioning
in a data center was designed to verify the optimization effect of the The empirical parameters in the heat transfer model for the
algorithm. The data center layout is shown in Fig. 8. The cooling cooling tower were obtained by using the least square method (95%
load model of the data center was established by DeST-c (com- confidence) using some experimental data. The fixed flow rates of
mercial). The outdoor dry-bulb temperature in this area obtained the cooling water and air were selected based on the test conditions
by DeST-c is shown in Fig. 9. The annual hourly cooling load of the of the model regression. Details of the working conditions of the
data center obtained by the simulation is shown in Fig. 10. regression test of the heat transfer model for the cooling tower are
The maximum hourly cooling load of the data center was shown in Table 3.
1401.49 kW, and the minimum hourly cooling load was 966.87 kW. The heat transfer capacity and the outlet water temperature of
The cooling load of the data center mainly came from the heat the cooling tower obtained by the test regression are calculated by
dissipation of the equipment in the room. Eq. (47) and Eq. (48), respectively.
The wet-bulb temperature was used as the environmental
parameter in the optimization model, so it was necessary to convert twi twb 2:690 ,t 1:099 m
twi
42 11 wb w
the hourly dry-bulb temperature and the moisture content into the h}wi hai
Qc ¼ (47)
wet-bulb temperature. To obtain the wet-bulb temperature, the twi twb 2:690 ,t 1:099 mw
1þ 42 11 twi wb ma
corresponding conversion program was given by using the rela-
tionship of the enthalpy. The calculation process is shown in Fig. 11.
2:690 ,t 1:099
twi twi twb
42 11
wb
two ¼ twi h}wi hai
4.3.2. Main equipment of the simulated case twb 2:690 ,t 1:099 , mw
cw 1 þ t42
wi
11 twi wb ma
According to the cooling load of the data center, the parameters
design of the main equipment in the cooling water system is shown (48)
in Table 2.
9
K. Ma, M. Liu and J. Zhang Energy 231 (2021) 120896
The experimental data were taken into the heat transfer model state. The comparison between the model predicted and experi-
for the cooling tower obtained by the regression, and the applica- mental values of the heat transfer capacity and the outlet water
bility of different flow rates of the cooling water and control mass temperature of the cooling tower are shown in Fig. 12 and Fig. 13
flow rates were analyzed under the non-regression parameter respectively.
Fig. 10. Year-round hourly cooling load results of the data center.
10
K. Ma, M. Liu and J. Zhang Energy 231 (2021) 120896
Table 3
Experimental conditions of the heat transfer model regression for the cooling tower.
Table 2
Main equipment parameters in the cooling water system.
11
K. Ma, M. Liu and J. Zhang Energy 231 (2021) 120896
Fig. 12. Comparison between the model predicted and experimental values of the heat transfer capacity of the cooling tower Qc.
Fig. 13. Comparison between the model predicted and experimental values of the outlet water temperature of the cooling tower two.
12
K. Ma, M. Liu and J. Zhang Energy 231 (2021) 120896
Table 5
Detail parameters of the equipment model.
13
K. Ma, M. Liu and J. Zhang Energy 231 (2021) 120896
Fig. 14. Distribution of the optimization result of total energy consumption under the HP-PSO algorithm.
Fig. 17. Variation of the optimized total energy consumption and the optimized outlet
Fig. 15. The minimum energy consumption of the system with different settings of the water temperature of the cooling tower with different cooling loads.
outlet water temperature of the cooling tower.
towers were regressed by experimental data. The characteristic
parameters of the heat transfer model for a cooling tower were
related to the heat transfer performance of the packing in the
cooling tower which had no relationship with the operating
state parameters.
2) Based on the heat transfer model for cooling towers, the online
operation optimization model for cooling water systems was
established. The HP-PSO algorithm was proposed to solve the
online operation optimization problems of continuous variables
and discrete variables in cooling water systems in cooling water
systems.
3) The HP-PSO algorithm could effectively find the optimum
running conditions. Compared with the T-PSO algorithm and
the RePSO algorithm, the HP-PSO algorithm had a higher
excellent rate and a lower degree of dispersion.
4) When the wet-bulb temperature was higher than 29 C, the
optimal value of the outlet water temperature of the cooling
tower increased to nearly 32 C. It was also shown that under
the condition of higher wet-bulb temperature, the setting of the
return water temperature difference of the cooling water at 5 C
was effective for energy-saving. When the wet-bulb tempera-
Fig. 16. Optimization results for different outdoor wet-bulb temperatures. ture was below 29 C, the setting of the return water tempera-
ture difference of the cooling water at 7 C was more energy-
saving.
14
K. Ma, M. Liu and J. Zhang Energy 231 (2021) 120896
Fig. 18. Comparison of energy consumption between the constant temperature difference algorithm and the HP-PSO algorithm under the mechanical cooling mode.
5) Compared with the optimization results of the constant tem- Pe shaft power (kW)
perature difference optimization algorithm, energy consump- DQe energy balance error (%)
tion of the cooling water system could be reduced by 15.3% with Qc heat transfer capacity of cooling tower (kW)
the use of the HP-PSO algorithm proposed in this paper. Qe heat transfer capacity of evaporator/cooling load (kW)
qc heat loss of condenser (kW)
qe heat loss of evaporator (kW)
Credit author statement S pipe network impedance (s2/m5)
t temperature ( C)
Keyan Ma: Conceptualization, Methodology, Software, Data Z height of the packing (m)
curation, Writing e original draft, Jili Zhang: Supervision, Writing-
Reviewing and Editing., Mingsheng Liu: Funding acquisition, Greek symbols
Visualization, Investigation. h efficiency (%)
l air-water mass ratio
Declaration of competing interest f relative humidity (%)
15
K. Ma, M. Liu and J. Zhang Energy 231 (2021) 120896
performance of counterflow wet-cooling tower. Energy 2014;74:855e62. [21] Ma Z, Wang S. Supervisory and optimal control of central chiller plants using
[9] Poppe M, Ro € gener H. Berechnung von rückkühlwerken. VDI-W€ armeatlas. simplified adaptive models and genetic algorithm. Appl Energy 2011;88:
1991. p. 1e15. 198e211.
[10] Jin G, Cai W, Lu L, Lee EL, Chiang A. A simplified modeling of mechanical [22] Seo J, Ooka R, Kim JT, Nam Y. Optimization of the HVAC system design to
cooling tower for control and optimization of HVAC systems. Energy Convers minimize primary energy demand. Energy Build 2014;76:102e8.
Manag 2007;48:355e65. [23] Chang Y. Optimal chiller loading by evolution strategy for saving energy.
[11] Asvapoositkul W, Treeutok S. A simplified method on thermal performance Energy Build 2007;39:437e44.
capacity evaluation of counter flow cooling tower. Appl Therm Eng 2012;38: [24] Ardakani AJ, Ardakani FF, Hosseinian SH. A novel approach for optimal chiller
160e7. loading using particle swarm optimization. Energy Build 2008;40:2177e87.
[12] Bornman W, Dirker J, Arndt DC, Meyer JP. Operational energy minimisation [25] Kusiak A, Li M, Tang F. Modeling and optimization of HVAC energy con-
for forced draft, direct-contact bulk air cooling tower through a combination sumption. Appl Energy 2010;87:3092e102.
of forward and first-principle modelling, coupled with an optimisation plat- [26] Beghi A, Cecchinato L, Cosi G, Rampazzo M. A PSO-based algorithm for optimal
form. Energy 2016;114:995e1006. multiple chiller systems operation. Appl Therm Eng 2012;32:31e40.
[13] Nasrabadi M, Finn DP. Performance analysis of a low approach low temper- [27] Lee K, Cheng T. A simulationeoptimization approach for energy efficiency of
ature direct cooling tower for high-temperature building cooling systems. chilled water system. Energy Build 2012;54:290e6.
Energy Build 2014;84:674e89. [28] Mu B, Li Y, Seem JE, Hu B. A multivariable Newton-based extremum seeking
[14] Huang S, Zuo W, Sohn MD. Improved cooling tower control of legacy chiller control for condenser water loop optimization of chilled-water plant. J Dyn
plants by optimizing the condenser water set point. Build Environ 2017;111: Syst Meas Contr 2015;137:1e10.
33e46. [29] Singh K, Das R. A feedback model to predict parameters for controlling the
[15] Yu FW, Chan KT. Optimization of water-cooled chiller system with load-based performance of a mechanical draft cooling tower. Appl Therm Eng 2016;105:
speed control. Appl Energy 2008;85:931e50. 519e30.
[16] Cortinovis GF, Paiva JL, Song TW, Pinto JM. A systemic approach for optimal [30] Keshtkar MM. Performance analysis of a counter flow wet cooling tower and
cooling tower operation. Energy Convers Manag 2009;50:2200e9. selection of optimum operative condition by MCDM-TOPSIS method. Appl
[17] Marques CAX, Fontes CH, Embiruçu M, Kalid RA. Efficiency control in a Therm Eng 2017;114:776e84.
commercial counter flow wet cooling tower. Energy Convers Manag 2009;50: [31] Gordon JM, Ng KC. Predictive and diagnostic aspects of a universal thermo-
2843e55. dynamic model for chillers. Int J Heat Mass Tran 1995;38:807e18.
[18] Chang C, Shieh S, Jang S, Wu C, Tsou Y. Energy conservation improvement and [32] Eberhart RC, Shi Y, Kennedy J. Swarm intelligence. Elsevier; 2001.
ONeOFF switch times reduction for an existing VFD-fan-based cooling tower. [33] Karami M, Wang L. Particle Swarm optimization for control operation of an
Appl Energy 2015;154:491e9. all-variable speed water-cooled chiller plant. Appl Therm Eng 2018;130:
[19] Chow TT, Zhang GQ, Lin Z, Song CL. Global optimization of absorption chiller 962e78.
system by genetic algorithm and neural network34; 2002. p. 103e9. [34] Abed KA, Khalil EE, Abouel-Fotouh AM, El-Hariry G, Abd El Salam LO. Optimal
[20] Lu L, Cai W, Soh YC, Xie L, Li S. HVAC system optimizationeecondenser water design of a counter flow cooling tower using PSO algorithm for operating cost
loop. Energy Convers Manag 2004;45:613e30. minimization. Appl Therm Eng 2018;143:149e59.
16