Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

Heat Mass Transfer (2013) 49:155–161

DOI 10.1007/s00231-012-1071-1

ORIGINAL

Natural draft dry cooling tower modelling


K. Tanimizu • K. Hooman

Received: 22 November 2011 / Accepted: 3 September 2012 / Published online: 18 September 2012
Ó Springer-Verlag 2012

Abstract Predictions based on a numerical simulation of m Mass flow rate


a natural draft dry cooling tower (NDDCT) has been p Pressure
compared with those obtained theoretically and experi- S The modulus of the mean-rate-of-strain tensor
mentally. Experiments are conducted in a lab-scale Q Internal heat generation rate
NDDCT and are validated with a three-dimensional t Time
numerical simulation of the flow in and around the heat T Temperature
exchangers, which is modelled as a porous medium. Both W Volume-averaged velocity
vertical and horizontal arrangements of the heat exchangers (u,v,w) Velocity components along (x,y,z) axes
are examined. The experimental, numerical and theoretical
Greek symbols
approaches lead to very close prediction for the air velocity
a Thermal diffusivity
and temperature at the exit of the cooling tower. Results of
q Density
this study are expected to be useful for future work on the
k Thermal conductivity
development of air-cooled condensers for geothermal
e Rate of energy dissipation
power plants in Australia.
b Thermal expansion coefficient
h Dimensionless temperature
List of symbols
u Porosity
l Viscosity
Variables
m Kinematic viscosity
A Cross section area
X Dimensionless parameter
CF Form drag coefficient
cp Specific heat Subscripts
d Diameter a Air
D Tower inlet/outlet opening ave Average
g Gravitational acceleration b Bundle/base
H Tower height e Tower exit
k Turbulent energy eff Effective
K Permeability f Fluid
l Finned-tube bundle thickness h Hydraulic
L Bundle height in Inlet
o Outlet
K. Tanimizu  K. Hooman (&) r Room
Queensland Geothermal Energy Centre of Excellence, s Solid phase
School of Mechanical and Mining Engineering,
t Turbulent
The University of Queensland, Brisbane,
QLD 4072, Australia T Tube
e-mail: k.hooman@uq.edu.au w Wall

123
156 Heat Mass Transfer (2013) 49:155–161

1 Introduction described by Odabaee and Hooman [14], with the main goal
of improving the cooling system for a geothermal power
Geothermal power plants are the major candidates of next plant. However, what is yet missing in the literature is a
generation renewable and emission-free power generation thorough knowledge about the scaling of cooling towers, for
systems in Australia. This technology is not currently at an example, one does not know the functional relation between
advanced stage of commercialization in Australia com- the height of the cooling tower and the heat that can be
pared to other renewable counterparts such as, hydro, wind dumped by different heat exchangers. That is, we do not
or solar. However, industry and renewable energy associ- know how the 2 m height model in our lab relates to a real
ations, together with government agencies, started includ- prototype which can be around 200 m; see [15]. Hence in
ing geothermal energy within their calculations of available the present work, we try to clarify the effect of the scaling of
generation capacity by targeting an annual renewable a tower using a laboratory scaled model (2 m height) and
energy generation of 45,000 GW by 2020 [1]. Geothermal later compare with a theoretical and a numerical model.
energy systems have the potential to produce a base load Further analysis of the flow structure inside the scaled
generation capacity capable of replacing existing coal-fired cooling tower is made based on the numerical model.
plants in the foreseeable future. However, there are some
technical challenges to overcome. One of the major tech-
nical difficulties is the design of cooling system used. 2 Theoretical analysis
Although wet cooling is more efficient than dry cooling,
water shortages and harsh environmental conditions in Initially, a theoretical analysis of the heat exchanger in the
areas such as the Australian desert have forced designers to cooling tower is made, assuming it to be a porous media.
consider less efficient and more expensive air-cooled sys- This simplistic approach is expected to give rough estimate
tems, or dry cooling as it is often termed. Air-cooled plants of the exit flow velocity to which the cooling tower per-
offer potential economic advantages due to plant sitting formance is related.
flexibility. Air cooling can be done by using fans In a NDDCT the driving force is the air density differ-
(mechanical driven) or by using natural draft through a ence which follows Boussinesq approximation [16] result
cooling tower. The fan-driven systems can be built quickly in
and at a relatively lower cost but their operating costs are Dp ¼ qgHbDT ð1Þ
higher due to their higher maintenance requirements and
various parasitic losses associated while running the fans Porous medium modelling of the heat exchanger leads to
[2]. The cooling system is a significantly costly item in the the following pressure drop for flow of air with a volume-
power plant and affects the performance of the entire averaged velocity W, across a finned-tube bundle of
power cycle. If the cooling system does not provide ade- thickness l
quate cooling, the overall plant efficiency decreases with  
lW CF qW 2
serious economic consequences (e.g. decreased electricity Dp ¼ l þ pffiffiffiffi ð2Þ
K K
production). It has been reported that approximately
0.3 GW per year of electrical generation in the United As shown by Hooman and Gurgenci [17] both form drag
States has been lost because of cooling towers operating coefficient and permeability change with internal flow
below their design efficiency. This corresponds to an eco- structure as well as the porosity of the porous medium.
nomic penalty of around 20 million US dollars per year [3]. Nonetheless, the average value of the form drag reported
It is therefore, very important to design and analyse highly there is O (10-1) while that of the permeability is O (10-5)
efficient dry cooling systems for power plants. Hence, in for a commercial finned-tube bundle. Hence, the criterion
order to improve the performance of cooling systems, for a form drag dominant flow through the bundle is
numerical or theoretical investigations have been reported W [ O ð1 cm/sÞ .
by many researchers ([4–11], to name a few) as reviewed The tower frictional pressure drop, for fluid velocity Wf,
by Kroger [12]. is the sum of distributed and local (changes in cross-sec-
On the other hand, few experimental investigations have tional area, recirculation, and other imperfections) losses
been reported for the fan-assisted NDDCTs (Natural Draft  
4H
Dry Cooling Towers), e.g. [10] and [13]. Since the size of a Dp ¼ 0:5qWf2 f þj ; ð3Þ
Dh
cooling tower is huge, it is very difficult to conduct on-site
experiments to collect data. In order to conduct research in a where j for our lab scale model, as given by Table 1.1 of
laboratory scale, the Queensland Geothermal Energy Centre Bejan and Lorente [19], puts on higher values than the
of Excellence (QGECE) tried to develop an efficient cool- other term in the parenthesis. The tower height and
ing system including a new type of heat exchanger as hydraulic diameter are, in most of the practical designs,

123
Heat Mass Transfer (2013) 49:155–161 157

comparable to each other so that one can simply neglect the


distributed losses. Then, the pressure drop through the
tower scales with local losses
Dp  0:5jqWf2 ð4Þ

The dimensionless total pressure drop should then scale


with
Dp
 Maxð1; XÞ ð5Þ
0:5jqWf2

The two pressure drop terms can be comparable when the


2
dimensionless group X ¼ 2lCpFffiffiuffi  Oð1Þ. For very high/low
j K
values of dimensionless parameter X, the tower/bundle
pressure drop is the dominant one. For the specific problem
of finned-tube bundle which is considered in Hooman and
Gurgenci [17], X  Oð10Þ. Thus the heat exchanger pres-
sure drop becomes the dominant one. However, for the
experiments conducted in a tower without heat exchangers, Fig. 1 Computational domain to simulate the lab scale model
as will be shown in the forthcoming discussion, the pres-
sure drop scales with the tower frictional losses. conducted using a commercially available CFD code
According to the first law of thermodynamics, the heat (ANSYS FLUENT). The computational domain, as
transferred to the fluid flowing through the porous medium depicted by Fig. 1, is a three-dimensional rectangular
increases the enthalpy of the fluid, i.e. Q ¼ qAWcp DT. This enclosure of size 3 m 9 5 m 9 4 m.
can be used to get the volume-averaged velocity as The laboratory scale of a natural draft dry cooling tower
Q system has been simplified to basic case of an isolated
W¼ ð6Þ
qAcp DT NDDCT structure. The atmospheric thermal conditions
surrounding the NDDCT have been presumed uniform with
Equation (6) is general enough to cover all heat exchanger
constant air temperature and pressure. A transient solution
configurations, however for comparison purpose,
is run with a time step of 1 s for a total duration of 30 min
horizontal and vertical bundle arrangements are further
with air being assumed as dry and incompressible. Radia-
examined in this paper. For the case of vertical
tion heat transfer is assumed to be negligible and a semi-
arrangement, the cross-sectional area A in Eq. (6) is
implicit method for pressure-linked equations (SIMPLE)
given by
algorithm is used for the calculation of pressure and con-
A ¼ pDh L; ð7Þ sequently the velocity field, which is needed for the solu-
tion of the energy equation. A total of 900,000 (hybrid
where for a horizontal tube arrangement, to a good unstructured) mesh elements and a first-order upwind dif-
approximation, the area is given by ferentiating scheme are used to approximate the spatial
p derivatives at all interior grid points. The k - e turbulence
A ¼ D2h ; ð8Þ
4 model is used as the fluid Rayleigh number based on the
room height is O (1010). Grid-independence was verified by
where Dh and L are the tower base hydraulic diameter and
control runs on a finer grid that produced consistent results
the heat exchanger height respectively. Equations (5, 6)
(with a maximum error being less than 3 %). Hence, finer
should be combined with Eqs. (7, 8) to give the fluid
grids were not used in reporting the results. It should be
velocity (and thus the mass flow rate) depending on the
noted that the convergence criterion (maximum relative
tower and heat exchanger design.
error in the values of the dependent variables between two
successive iterations) in all runs was set at 10-5.
Adiabatic and no-slip wall boundary conditions were used
3 Numerical analysis to bound fluid and solid regions. For a previous investigation,
Hooman and Gurgenci [17], indicated that the porous med-
In order to understand the flow structure occurring inside ium formalism could be used to model a conventional finned-
the cooling tower, a detailed numerical analysis is tube heat exchanger bundle. Hence, in the present study, the

123
158 Heat Mass Transfer (2013) 49:155–161

same porous medium model is used where, the permeability in the porous region where in non-porous region one
K is calculated by using the Carman–Kozney correlation. simply uses u = 1 to recover fluid property.
dT2 u3
K¼ ; ð9Þ
100ð1  uÞ2 4 Experimental analysis
The form drag coefficient used in this study is given by
In order to validate the theoretical and numerical model
CF ¼ 0:55½9:887ð1  uÞðu  0:323Þ  0:8443: ð10Þ
used in the study, a simple experimental rig is constructed
The porous zone (heat exchangers) at the bottom of the to measure the exit flow velocities from a scaled cooling
tower forms either a horizontal layer or a vertical layer tower. Figure 2 shows the small scale cooling tower and
surrounding the tower base with a uniform and constant heating element that was used in this study. The cooling
heat generation rate fixed at 1.0 kW. The initial room tower consisted of the tower shell that was made of poly-
temperature and pressure are set at 300 K and 101,325 Pa, carbonate, the tower support, and the electric heating ele-
respectively. ment. Due to manufacturing difficulties, the cooling tower
The governing equations as used in the calculation are was made into a pyramidal-square shape.
represented below in tensor notation [18] The cooling tower shell was supported by four bars each
having a height of 30 cm. The base area was 2 m2 with the
oui tower exit area being D2o (0.92 m 9 0.92 m). Four copper
¼ 0; ð11Þ heating bars, 10 mm diameter, were arranged horizontally
oxi
in parallel to form the heating element. Static temperature
oðui Þ oðui uj Þ 1 op
þ ¼ measurements were performed as a shortcut to predict the
ot oxj q oxi    velocity which was very hard to measure directly. The
o oui ouj 2
þ ð m þ mt Þ þ  dij k power consumed by the heating element was measured by
oxj oxj oxi 3pffiffiffiffi using ‘‘Nanovip power meter’’. During the experiment, the
m þ m t þ C F j ui j K
þ! g bDT  Fui room temperature was measured using a ‘‘tech’’ ther-
K mometer (NO.CN-306, K-type thermocouples). The sur-
ð12Þ face temperature on the copper tube was measured by a k-
 
ok oðkui Þ o ok pffiffiffiffiffiffiffiffiffiffiffiffi type thermocouple and was controlled by an in-house
þ ¼ ðm þ 0:77mt Þ þ mt 2Sij Sij manufactured temperature control box. The room temper-
ot oxi oxj oxj
ð13Þ ature and temperatures at the tower inlet and outlet were
gi oq
 2 e measured using four ‘‘Go! Temp’’ temperature probes (see
q oxi
  Fig. 3) so that the tower exit velocity could be estimated
oe oðeui Þ o oe using Eq. (6). Next section presents the results of the
þ ¼ ð m þ mt Þ
ot oxi oxj  oxj analysis made based on the observations, which includes
1:44mt e pffiffiffiffiffiffiffiffiffiffiffiffi v  g oq 
  i uncertainty of measurements [20].
þ 2Sij Sij  tanh  2
k u q oxi
1:92e2

k
ð14Þ
 
oT oðuj TÞ o mt oT
þ ¼ aeff þ
ot oxj oxj 0:9 oxj
Q
þF ; ð15Þ
u qcp f þð1  uÞ qcp s

where F is a binary factor equal to unity in the porous region


and zero elsewhere. Furthermore, S is the modulus of the
mean-rate-of-strain tensor and mt = 0.09 k2/e (in porous
layer: k = 1.5u2 (u2 ? v2)1/2/104 and e = 1.643 k3/2).
Note that the effective thermal diffusivity is given by

ukf þ ð1  uÞks
aeff ¼ ð16Þ
u qcp f þð1  uÞ qcp s
Fig. 2 Experimental setup

123
Heat Mass Transfer (2013) 49:155–161 159

Porosity = 0.66
Porosity = 0.70
0.5 Porosity = 0.74
Vertical
Porosity = 0.78

0.4
Horizontal

0.3

w/(2W)
0.2

Fig. 3 Thermocouple locations to measure the tower exit tempera- 0.1


ture (mid-plane line, of length Do, is shown)

0
Table 1 Comparison between numerical prediction and experimental 0 0.1 0.2 0.3 0.4 0.5
data in horizontal arrangement x/D o
Data Scaling Experiment CFD
Fig. 4 Normalized velocity profiles at the tower exit mid-plane; W is
Mean exit velocity 0.52 0.54 (±0.064 m/s) 0.47 the average velocity in the cross-section

5 Results and discussions for the horizontal case.


Figure 4 illustrates local normalized (upward) velocity
Two different arrangements for heat exchanger bundles are distribution on the tower exit mid-plane for both vertical
numerically examined in this study, which is, horizontal and horizontal heat exchanger arrangements. The heat
and vertical. However, prior to investigating the effects of exchanger porosity is systematically varied from 0.66 to
the heat exchanger arrangement, a comparison between 0.78 following Hooman and Gurgenci [17]. As can be seen,
numerical results and theoretical prediction as well as increasing the porosity leads to a higher velocity in both
experimental data can be made for validation purpose. The cases. Interestingly, higher velocities are observed in ver-
experiments using the laboratory scale model were only tical arrangement than in a horizontal arrangement. It is
conducted in horizontal arrangement. The power of heater mainly due to the fact that the driving force, the chimney
was kept at 1 kW. Tower inlet and outlet temperature were effect, is linearly proportional to the effective height, which
measured during the experiments. The measured mean is higher for a vertical heat exchanger. The effective height
velocity at the tower exit plane was then estimated from for a vertical bundle is measured from the tower opening
Eq. (6). Table 1 shows the comparison between the mea- (half way through) to the tower exit. For the vertical case
surement data and numerical result of the air velocity at the the effective height is estimated to be 2.15 m whereas for
tower exit plane. It is observed that while the experimental the horizontal case its value is 2 m. This indicates that, to
data for the tower exit velocity is close to 0.54 m/s the first approximation, one can expect the average velocity
(±0.064 m/s), the result of our theoretical calculation for the vertical case to be almost 8 % higher compared to
agrees with the measurement value within 5 % of this that of horizontal case. This also indicates that, in the
velocity, [16]. At the same time, the numerical prediction absence of atmospheric disturbances, the amount of heat
for the mean tower exit velocity is evaluated to be equal to rejected by a particular tower where the heat exchangers
0.47 m/s. The numerical value is slightly under predicted. are arranged vertically around the tower inlet shows better
It should be noted that, for experimental evaluation, instead cooling performance than that ejected by the same tower
of Eqs. (7, 8), the following equations were used for the with the heat exchangers arranged horizontally in the inlet
cross-sectional areas: cross-section.
A ¼ Din L ð17Þ Figure 5 indicates the dimensionless temperature profile
on the same mid-plane as that of Fig. 2 for the aforemen-
for a vertical arrangement and
tioned range of porosity. The dimensionless temperature is
A ¼ D2in ð18Þ defined as h = (T - Twall)/(Twall - Tini) with Twall and Tini

123
160 Heat Mass Transfer (2013) 49:155–161

Porosity = 0.66
Porosity = 0.70
0.045 Porosity = 0.74
Porosity = 0.78

0.035
Horizontal
Vertical

0.025
θ

0.015

0.005

0.1 0.2 0.3 0.4 0.5


x/D o

Fig. 5 Dimensionless temperature profiles at the tower exit for both


vertical and horizontal case

being the wall temperature (at the tower exist) and initial
room temperature, respectively. The temperature distribu-
tion is observed to be flatter for the horizontal case, similar
to the velocity distribution.
As a sample of our results, Figs. 6 and 7 are presented to
indicate the velocity vector plots (coloured by velocity
magnitude) and dimensionless isotherms for both hori-
zontal and vertical bundle arrangements with u = 0.66 on
x = 0 plane. As can be observed, the flow patterns near and
inside the cooling tower are different due to the arrange-
ments of the bundles. Flow separation occurs at the lower
Fig. 6 Velocity vector plots on x = 0 mid-plane for a horizontal and
edge of the tower wall (at the corner of inlet). The sepa- b vertical bundle arrangement (u = 0.66)
ration formed a vena contracta with a corresponding dis-
torted inlet velocity distribution. After the flow passes the
heat exchanger bundles, two recirculation zones are formed Figure 7a, b illustrate the dimensionless temperature
on either side of the tower, where the air velocity is much distributions for the same porosity and bundle arrange-
lower when compared to the regions close to the tower ments as those of Fig. 6, again on x = 0 mid-plane. Note
centre. With the vertical bundles, the flow tends to be more that the average wall temperature at the tower exit is used
or less uniform after exiting the heat exchangers. Another as reference temperature. Hence, the negative values for h
interesting observation is that, the maximum velocity is can be explained as regions where the air temperature is
higher with the vertical case and the velocity distribution at less than that of the wall at the tower exit. In horizontal
the tower entrance is different as one would expect case, the distribution of temperature is almost uniform
(because of different flow resistances; with vertical bundles except near bottom corner of the tower. The two near-wall
the air faces heat exchanger resistance at the tower inlet recirculation regions for the horizontal case result in high
whereas with the other case, this flow resistance occurs temperature regions inside the tower, which is more or less,
further downstream in the tower). In both cases, vertical symmetric to each other. These hot spots however could
and horizontal heat exchangers, the air is heated and have been avoided by a better design for the cooling tower.
pushed up towards the adiabatic horizontal wall. As the This fact brings home the point that tower design is not
walls are adiabatic, the heat transferred to air by the heat only dictated by external effects, say wind, but also by
sources (bundles), keeps on increasing the air temperature minimizing the inside pressure drops and eliminating the
until thermal equilibrium is reached. circulations inside the tower. In a vertical case, temperature

123
Heat Mass Transfer (2013) 49:155–161 161

examining different heat exchangers in the cooling tower


and study the effects of cross-wind on the heat and fluid
flow from the air-cooled heat exchangers.

References

1. Change DoC (2009) Renewable energy target. http://www.climate


change.gov.au/en/government/initiatives/renewable-target.aspx
2. Hooman K, Gurgenci H (2010) Different heat exchanger options
for natural draft cooling towers. World Geothermal Congress
Nusa Dua, Bali, Indonesia. Paper # 2633
3. Tanimizu K, Hooman K (2010) Scaling laws for a natural draft
cooling tower: porous medium modeling of the heat exchangers.
In: Vafai K (ed) 3nd international conference on porous media
and its applications in science and engineering, Montecatini, Italy
4. Al-Waked R (2009) Development of performance-improving
structures for power station cooling towers. University of New
South Wales, Sydney
5. Al-Waked R, Behnia M (2004) The performance of natural draft
dry cooling towers under crosswind: CFD study. Int J Energy Res
28(2):147–161
6. Al-Waked R, Behnia M (2005) The effect of windbreak walls on
the thermal performance of natural draft dry cooling towers. Heat
Transf Eng 26(8):50–62
7. Dupreez AF, Kroger DG (1993) Effect of wind on performance of
a dry-cooling tower. Heat Recovery Syst CHP 13(2):139–146
8. Dupreez AF, Kroger DG (1995) Effect of the shape of the tower
supports and walls on the performance of a dry-cooling tower
subjected to cross winds. Heat Transf Eng 16(2):42–49
9. Furzer IA (1968) Natural draft cooling tower—an approximate
solution. Ind Eng Chem Process Design Dev 7(4):555–560
10. Zhai Z, Fu S (2006) Improving cooling efficiency of dry-cooling
Fig. 7 Dimensionless temperature distributions for a horizontal and towers under cross-wind conditions by using wind-break meth-
b vertical bundle arrangement ods. Appl Therm Eng 26(10):1008–1017
11. Fu S, Zhai ZQ (2001) Numerical investigation of the adverse
effect of wind on the heat transfer performance of two natural
field is also symmetric with no hot spots observed inside draft cooling towers in tandem arrangement. Acta Mech Sin
the tower, and the temperature gradually decreases with 17(1):24–34
height. 12. Kroger DG (2004) Air-cooled heat exchangers and cooling
towers. Pennwell Corp., Tulsa
13. Wei QD, Zhang BY, Liu KQ, Du XD, Meng XZ (1995) A study
of the unfavorable effects of wind on the cooling efficiency of dry
6 Conclusions cooling-towERS. J Wind Eng Ind Aerodyn 54:633–643
14. Odabaee M, Hooman K (2012) Metal foam heat exchangers for
Three-dimensional numerical simulation of the perfor- heat transfer augmentation from a tube bank. Appl Therm Eng
36:456–463
mance of a small scale NDDCT is presented and compared 15. Busch D, Harte R, Kratzig WB, Montag U (2002) New natural
with both experimental and theoretical data. Simulation draft cooling tower of 200 m of height. Eng Struct 24(12):
results indicate that the porous medium modelling is suit- 1509–1521
able in predicting the flow in and around the heat 16. Hooman K (2010) Dry cooling towers as condensers for geo-
thermal power plants. Int Commun Heat Mass Transf 37(9):
exchanger bundle. In addition, the k - e turbulence model 1215–1220
is found to be a reliable model to predict the complex 17. Hooman K, Gurgenci H (2010) Porous medium modeling of air-
nature of the flow in and around a cooling tower despite the cooled condensers. Transp Porous Media 84:257–273
fact that the model is believed to deviate from experimental 18. ANSYS FLUENT 12.0 Theory Guide
19. Bejan A, Lorente S (2008) Design with constructal theory. Wiley,
results for low fluid velocities. The method presented in Hoboken
this paper makes it possible to conduct further research on 20. Mee DJ (1993) Uncertainty analysis of conditions in the test
the design and performance of NDDCTs without the need section of the T4 shock tunnel. Division of Mechanical Engi-
of complex engineering details or iterative solution to the neering, The University of Queensland, Brisbane
draft equation. Future research at QGECE will aim at

123

You might also like