See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/221914103

Optimal Design of Cooling Water Systems

Chapter · August 2011

DOI: 10.5772/18894 · Source: InTech

CITATIONS READS
23 1,728

3 authors, including:

Jose Maria Ponce-Ortega

Universidad Michoacana de San Nicolás de Hidalgo
378 PUBLICATIONS   5,989 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Reuse of water in a residential complex View project

The 13th Conference on Sustainable Development of Energy, Water and Environment Systems – SDEWES2018 View project

All content following this page was uploaded by Jose Maria Ponce-Ortega on 28 May 2014.

The user has requested enhancement of the downloaded file.

 8

Optimal Design of
Cooling Water Systems
Eusiel Rubio-Castro, José María Ponce-Ortega
and Medardo Serna-González
Chemical Engineering Department, Universidad Michoacana de San Nicolás de Hidalgo
México

1. Introduction
Re-circulating cooling water systems are generally used to remove waste heat from hot
process streams in conditions above the ambient temperature in many types of industries
such as chemical and petrochemical, electric power generating stations, refrigeration and
air conditioning plants, pulp and paper mills, and steel mills. Typical re-circulating
cooling water systems are constituted by a mechanical draft wet-cooling tower that
provides the cooling water that is used in a set of heat exchangers operated in parallel as
can be seen in Figure 1. The economic optimization of re-circulating cooling water
systems includes the simultaneous selection of the optimal design variables of the cooling
tower and each heat exchanger in the cooling network, as well as the optimal structure of
the cooling water network. The question is then how to reach this goal. Earlier work on
cooling water systems has concentrated on the optimization of stand-alone components,
with special attention given on the individual heat exchangers of the cooling water
network. Other publications have dealt with the problem of designing minimum-cost
cooling towers for a given heat load that must be dissipated (see Söylemez 2001, 2004;
Serna-González et al., 2010). Most of the methodologies previously reported have
concentrated their attention in the optimal synthesis of cooling water networks (see Kim
and Smith, 2001; Feng et al., 2005; Ponce-Ortega et al., 2007). All previous formulations
simplified the network configurations because they consider the installation of only one
cooling tower; however, the industrial practice shows that sometimes it is preferable to
use a set of cooling towers connected in series, parallel, and series-parallel arrangements
to improve the performance of the cooling towers reducing the operational cost and,
hence, to decrease the overall total annual cost for the cooling water system. In addition,
previous methodologies do not have considered several arrangements for the cooling
water that can improve the performance in the coolers and reduce their capital costs.
Another limitation for the previously reported methodologies is that they are based on the
use of simplified formulations for the design of cooling towers.
This chapter presents an optimization model for the simultaneous synthesis and detailed
design of re-circulating cooling water systems based on the superstructure of Figure 2. The
model considers all the potential configuration of practical interest and the results show the
significant savings that can be obtained when it is applied.
162 Energy Management Systems

Fig. 1. Typical cooling water system.

2. Model formulation
This section presents the relationships for the proposed model, which is based in the
superstructure of Figure 2. In the next equations, the set NEF represents the cooling medium
streams leaving the cooler network, ST the stages of the cooling network, HP the hot process
streams and NCT represents the cooling towers. The subscripts av, b, cu, d, dis, ev, f, fi, fr, i, j, k, e,
l, m, in, n, nct , out, p, pl, r, s, t and WB are used to denote average, blowdown, cooling
medium, drift, end of the cooling tower network, evaporation, fan, fill, cross-sectional, hot
process stream, cold process stream, stage in the cooling network, type of packing, constants to
calculate the heat and mass transfer characteristics for a particular type of packing, constants
for the loss coefficient correlation for a particular type of packing, inlet, temperature increment
index, cooling tower, outlet, pump, parallel arrangement, makeup, series arrangement, total
and wet-bulb, respectively. The superscript max is an upper limit and min is a lower limit. In
addition, the scalars NOK is the total number of stages in the cooling network, NCP is the total
number of cooling medium streams at the hot end of the cooling network and LCT is the last
cooling tower in the cooling tower network.
The heat of each hot process stream (QHPi) is calculated by the multiplication between the
heat capacity flowrate of each hot process stream (FCPi) and the difference of the inlet and
outlet temperatures of each stream (THINi, THOUTi), All terms of the above equation

THINi  THOUTi  FCPi  QHPi (1)

Optimal Design of Cooling Water Systems 163

Fig. 2. Proposed superstructure.

QHPi   qi , k , i  HP (2)
kST

are parameters with known values. The heat absorbed by the cooling medium in the
matches is equal to the transferred heat by the hot process streams, where qi,k is the heat
exchanged in each match. In addition to the above balance, the model includes balances for
the splitters and mixers on each stage for the cooling medium.
A heat balance for each match of the superstructure is required to determine the
intermediate temperatures of the hot process stream and the cooling medium as well as the
cooling medium flowrate at each match. For problems with NH hot process stream, the
number of stages in the superstructure NOK must be equal to NH, to allow arrangements
completely in series. In this case, there are NOK+1 temperatures for each hot process stream,
because the outlet temperature at one stage is equal to the inlet temperature in the next
stage. It is required to identify the inlet and outlet temperatures at each stage for the cooling
medium. Stage k=1 represents the lowest level of temperature. The heat balance for each
match is given as follows:

Thi , k  1  Thi , k  FCPi  qi , k , k  ST , i  HP (3)

Tcouti , k  Tcink  Fi , k  CPcu  qi , k , k  ST , i  HP (4)

 
164 Energy Management Systems

where Fi,k is the flowrate for the cooling medium in each match. In addition, Thi,k is the
temperature of each hot stream in each match, Tcink is the inlet temperature for the cooling
medium in each stage, Tcouti,k is the outlet temperature for the cooling medium in each
match and CPcu is the heat capacity for the cooling medium. In previous equations, CPcu is a
parameter known prior to the optimization process.
Mass and energy balances are required to calculate the inlet flow rate and temperature for
the cooling water to each stage. For the kth stage of the cooler network, the mass balance in
the mixer is given by the sum of the water flow rates that are required by the matches in that
stage ( i
F
HP
i ,k
) and the supply cold-water flow rate to the hotter adjacent stage k+1 (FFk+1)
minus the sum of the bypass water flow rate (FOk) running from the splitter of that stage to
the cooler network outlet and the flow rates of cooling water streams that are required by
the matches of the hotter adjacent stage k+1 ( iHP
Fi ,k 1 ),

 Fi ,k  FFk  1  FOk   Fi , k  1 , k  ST  1 (5)

iHP iHP

To calculate the temperature of the bypass cooling-water stream of each stage (TOk), the
following heat balance in the splitters is required,

 
  Fi , k  TOk    Fi , kTcouti , k  , k  ST  1 (6)
 iHP  iHP

Note that using the heat balance given in equation (4) is possible to know the inlet cooling
medium temperature ( Tcink  1 ) to each stage, except to the first one (i.e., for k = 1). And for
the first stage, the cooling medium temperature is given by the cooling tower,

 
  Fi , k  TOk  FFk  1TCUin  FOkTOk    Fi , k  1Tcink  1  , k  ST  1 (7)
 iHP  iHP

The above set of equations is necessary for NOK-1 stages.

The following mass balance must be included for the first stage to determine the flowrate of
the cold water provided by the cooling tower network to the cooler network, considering
that only cold water is used in the first stage.

FFk   Fi , k , k1 (8)

iHP

The inlet temperatures of the hot process streams define the last location for the
superstructure. In other words, the inlet temperature of the hot process stream i is the
temperature of such stream in the hot end of the cooling network (Thi,NOK+1),

THIN i  Thi , NOK  1 , i  HP (9)

The outlet temperatures of hot process streams give the first location for the superstructure.
Therefore, the outlet temperature of the hot process stream i is the temperature of such
stream in the cold end of the cooling network (Thi,1),
Optimal Design of Cooling Water Systems 165

THOUTi  Thi ,1 , i  HP (10)

In addition, the inlet temperature for the cooling medium (TCUIN) represents the inlet
temperature at the first stage (Tcin1), considering that the outlet temperature in each match
max
is restricted by an upper limit ( Tcout i ,k
) to avoid operational problems.

TCUIN  Tcin1 (11)

max
Tcouti , k  Tcout i ,k
, i  HP , k  ST (12)

To ensure a monotonically decrement for the temperatures through the stages of the
superstructure, the next constraints are included. It is necessary to specify that the
temperature of each hot process stream in the stage k must be lower or equal than the
temperature of each hot process stream in the stage k+1,

Thi , k  Thi , k  1 , k  ST , i  HP (13)

The inlet cooling medium temperature to stage k must be lower or equal than the outlet
cooling medium temperature in the match i, k,

Tcink  Tcouti , k , k  ST , i  HP (14)

The temperature inlet cooling medium to the cooling network must be less or equal than the
inlet cooling medium temperature in the stage k,

TCU in  Tcink , k  ST , k  1 (15)

Finally, the outlet cooling medium temperature in the match i,k should be lower or equal
than the outlet cooling medium temperature in the match i,k+1,

Tcouti , k  Tcouti , k  1 , k  ST , i  HP (16)

Logic constraints and binary variables are used to determine the existence of the heat
exchangers between the hot process stream i in the stage k with the cooling medium. These
constraints are stated as follow,

qi , k  max 1
qi ,k zi , k  0, i  HP , k  ST
(17)

Here max 1
qi ,k is an upper limit equals to the heat content of the hot process stream i and zi , k is

a binary variable used to determine the existence of the heat exchangers.

Because the area requirements for each match (Ai,k) are included in the objective function,
the temperature differences should be calculated. The model uses a pair of variables for the
temperature difference in the cold side (dtfrii,k) and the hot side (dtcali,k) of each match. In
addition, binary variables are used to ensure positive temperature differences and greater
than a given value of TMIN when a match exists.

 
dtcali , k  Thi , k  1  Tcouti , k   i 1  zi1, k , k  ST , i  HP
(18)
166 Energy Management Systems

 
dtfrii , k  Thi , k  Tcink   i 1  zi1, k , k  ST , i  HP (19)

dtcali , k  TMIN , k  ST , i  HP (20)

dtfrii , k  TMIN , k  ST , i  HP (21)

where  i is an upper limit for the temperature difference for the hot process stream i. The
value of  i is a constant known previous to the optimization and it is given by,

 i  max  0, THIN i  TCUIN , THOUTi  TCU in , THIN i  Tcoutmax , THOUTi  Tcoutmax  (22)

Equations (18) and (19) are written as inequalities because the heat exchanger costs decrease
when the temperature differences increase. Note that the use of binary variables allows the
feasibility because if the match does not exist, the parameter  i ensures that these
restrictions are met. When a heat exchanger for the hot process stream i exists at the stage k,
the binary variable zi1, k is equal to one, then the constraint is applied and the temperature
differences are properly calculated.
The flowrates (Fwj) and temperatures (Twj) of the cooling water streams that are directed to
the splitters at the inlet of the cooling tower network are determined in the last stage of the
cooler network. It is important to note that a problem with NH hot process streams will have
NH+1 cooling water streams leaving the cooler network, because in addition to the
flowrates of the cooling water for each match in the last stage it is generated an overall
cooling water stream that results from combining the bypass cooling-water streams of the
previous stages. Therefore, the value for the set of cooling water streams directed to the
cooling tower network is NEF=NH+1.

Fw j  Fi , k , k  NOK ; j  NCP ; j  i
(23)

Fw j   FOk , j  NCP (24)

kST

Tw j  Tcouti , k , k  NOK ; j  NCP ; j  i

(25)

Tw j Fw j   TOkFOk , j  NCP (26)

kST

The outlet cooling water stream flowrate from the cooling network can be sent to each tower
of the cooling tower network (Fw1,j,net) and/or directed to the end of the cooling tower
network (Fw2,j),

Fw j   Fw1, j ,nct  Fw2, j , j  NEF (27)

nctNCT

the inlet water flowrate (Fwin,net) and temperature (Twin,net) to the cooling towers are
generated by the mix of the portions of cooling medium streams sent to the cooling towers
and the flowrate from the cooling tower (FTTnet-1,net),

Fwin ,nct   Fw1, j ,nct  FTTnct 1,nct , nct  NCT (28)

jNEF
Optimal Design of Cooling Water Systems 167

Twin ,nct Fwin , nct   Tw j Fw1, j ,nct  Twout , nct  1FTTnct  1,nct , nct  NCT (29)
jNEF

There is a loss of water in the cooling towers by evaporation (Fwev,net) and the drift of water
by the air flowrate (Fwd,net). The water evaporated is obtained from the following
relationship:

Fwev ,nct  Fanct  wout , nct  win ,nct  , nct  NCT (30)

while the drift loss of water is 0.2 percent of the inlet water flowrate to the cooling tower
(Kemmer, 1988):

Fwd , nct  0.002Fwin , nct , nct  NCT (31)

Thus, for each cooling tower, the outlet water flowrate is given by the following expression:

Fwout ,nct  Fwin ,nct  Fwev ,nct  Fwd ,nct (32)

and the outlet cooling tower flowrate (Fwout,net) can be split and sent to the next cooling
tower and/or to the end of the cooling tower network (Fwsnet),

Fwout ,nct  FTTnct ,nct  1  Fwsnct , nct  NCT (33)

The flowrate (Fwdis) and temperature (Twdis) for the end of the cooling tower network are the
sum of the bypassed flowrates of the cooling medium streams and the outlet cooling tower
flowrates,

Fwdis   Fw2, j   Fwsnct (34)

jNEF nctNCT

TwdisFwdis   Tw j Fw2, j   Twout ,nct Fwsnct (35)

jNEF nct 1NCT

To avoid salts deposition, usually a little blowdown flowrate (Fwb) is applied over the water
flowrate treated in the cooling tower network, which can be determined by,

Fwr
Fwb    Fwd ,nct (36)
NCYCLES nctNCT

Note that the last term is the total drift loss of water by the air in the cooling tower network.
Also, Fwr is the makeup flowrate and NCYCLES is the number of concentration cycles. Then,
the outlet flowrate of the cooling tower network (Fwctn) is equal to the flowrate at the end of
cooling network minus the blowdown,

Fwctn  Fwdis  Fwb (37)

but the cooling medium temperature in the outlet of cooling network (Twctn) is,

Twctn  Twdis (38)

168 Energy Management Systems

To maintain the cooling medium flowrate constant in the cooling system, it is necessary a
makeup flowrate to replace the lost water by evaporation, drift and blowdown,

Fwr   Fwev ,nct   Fwd , nct  Fwb (39)

nctNCT nctNCT

Note that the total water evaporated and drift loss of water in the cooling tower network are
considered. The flowrate required by the cooling network (FCUin) is determined as follow:

FCU in  Fwctn  Fwr (40)

and the inlet cooling medium temperature to the cooling network is obtained from,

TCU inFCU in  TwctnFwctn  Twr Fwr (41)

To avoid mathematical problems, the recycle between cooling towers is not considered;
therefore, it is necessary to specify that the recycle in the same cooling tower and from a
cooling tower of the stage nct to the cooling tower of stage nct-1 is zero,

FTTnct 1,nct  0, nct , nct 1  NCT ; nct 1  nct (42)

The following relationships are used to model the design equations for the cooling towers to
satisfy the cooling requirements for the cooling network. First, the following disjunction is
used to determine the existence of a cooling tower and to apply the corresponding design
equations,

 2
znct 
 max
   z2 
nct
  nct   nct     , nct  NCT
 min  
  nct  0 
  nct   nct 
2
Here ZNCT is a Boolean variable used to determine the existence of the cooling towers,
 nct is an upper limit for the variables,  min
max
nct is a lower limit for the variables,  nct is any
design variable of the cooling tower like inlet flowrate, mass air flowrate, Merkel number,
and others. For example, when inlet flowrate to the cooling tower is used, previous
disjunction for the inlet flowrate to the cooling tower is reformulated as follows:

max
Fwin ,nct  Fw z1  0, nct  NCT
in ,nct nct
(43)

min
Fwin ,nct  Fw z1  0, nct  NCT
in ,nct nct
(44)

where max min

Fwin ,nct and  Fwin ,nct are upper and lower limits for the inlet flowrate to the cooling
tower, respectively. Notice that this reformulation is applied to each design variable of the
cooling towers. The detailed thermal-hydraulic design of cooling towers is modeled with
Merkel’s method (Merkel, 1926). The required Merkel’s number in each cooling tower, Menct,
is calculated using the four-point Chebyshev integration technique (Mohiudding and Kant,
1996),
Optimal Design of Cooling Water Systems 169

4
Menct  0.25CPcu Twin ,nct  Twout , nct   1 hn ,nct ; nct  NCT (45)
n1

where n is the temperature-increment index. For each temperature increment, the local
enthalpy difference ( hn ,nct ) is calculated as follows

hn ,nct  hsan ,nct  han ,nct , n  1,..., 4; nct  NCT (46)

and the algebraic equations to calculate the enthalpy of bulk air-water vapor mixture and
the water temperature corresponding to each Chebyshev point are given by,

CPcuFwin ,nct
han ,nct  hain ,nct 
Fanct
Twn ,nct  Twout ,nct  , n  1,..., 4; nct  NCT (47)

Twn ,nct  Twout ,nct  TCH n Twin ,nct  Twout , nct  , n  1,..., 4; nct  NCT (48)

where TCHn is a constant that represents the Chebyshev points (TCH1=0.1, TCH2=0.4,
TCH3=0.6 and TCH4=0.9). The heat and mass transfer characteristics for a particular type of
packing are given by the available Merkel number correlation developed by Kloppers and
Kröger (2005):

c 2 ,nct c 3 ,nct
 Fw   Fa 
 L fi ,nct 
1  c 4 ,nct
Twin ,nct 
c 5 ,nct
Menct  c1, nct  in ,nct   nct
 , nct  NCT (49)
 A fr ,nct   A fr ,nct 
   
To calculate the available Merkel number, the following disjunction is used through the
e
Boolean variable Ynct :

 1
Ynct   2
Ynct   3
Ynct 
     
  splash fill    trickle fill    film fill   , nct  NCT
 1   2   3 
cl ,nct  cl , nct , l  1,..., 5  cl ,nct  cl ,nct , l  1,..., 5  c l ,nct  cl ,nct , l  1,..., 5 
Notice that only when the cooling tower ntc exists, its design variables are calculated and
only one fill type must be selected; therefore, the sum of the binary variables referred to the
different fill types must be equal to the binary variable that determines the existence of the
cooling towers. Then, this disjunction can be described with the convex hull reformulation
(Vicchietti et al., 2003) by the following set of algebraic equations:
1 2 3 2 (50)
ynct  ynct  ynct  znct , nct  NCT

cl ,nct  cl1,nct  cl2,nct  cl3, nct , l  1,..., 5; nct  NCT (51)

cle,nct  ble ynct

e
, e  1,..., 3.; l  1,..., 5; nct  NCT (52)

Values for the coefficients ble for the splash, trickle, and film type of fills are given in Table 1
(Kloppers and Kröger, 2005); these values can be used to determine the fill performance. For
170 Energy Management Systems

each type of packing, the loss coefficient correlation can be expressed in the following form
(Kloppers and Kröger, 2003):

  Fw
d
 2 ,nct  Fa  3 ,nct
d
 Fw  5 ,nct  Fa  6 ,nct 
d d

K fi ,nct   d1,nct  in ,nct   nct   d4,nct  in ,nct   nct  L , nct  NCTa (53)
  A fr ,nct   A fr ,nct   A fr ,nct   A fr ,nct   fi ,nct
         

The corresponding disjunction is given by,

 1
Ynct   2
Ynct   3
Ynct 
     
  splash fill    trickle fill    film fill   , nct  NCT
 1   2   3 
 dm ,nct  dm ,nct , m  1,...,6   dm ,nct  dm ,nct , m  1,...,6   dm ,nct  dm ,nct , m  1,...,6 
Using the convex hull reformulation (Vicchietti et al., 2003), previous disjunction is modeled
as follows:

dm ,nct  dm1 ,nct  dm2 ,nct  dm3 , nct , m  1,...,6; nct  NCT (54)

dme ,nct  cme ynct

e
, e  1,..., 3; m  1,...,6;nct  NCT (55)

ble
l e=1 e=2 e=3
(splash fill) (trickle fill) (film fill)
0.249013 1.930306 1.019766
2 -0.464089 -0.568230 -0.432896
3 0.653578 0.641400 0.782744
4 0 -0.352377 -0.292870
5 0 -0.178670 0
Table 1. Constants for transfer coefficients

Values for the coefficients cme for the three fills are given in Table 2 (Kloppers and Kröger,
2003). These values were obtained experimentally and they can be used in the model
presented in this chapter. The total pressure drop of the air stream is given by (Serna-
González et al., 2010),

2
Favav
Pt ,nct  0.8335 , nct
 av ,nct A2fr ,nct
 K fi ,nctL fi ,nct  6.5 , nct  NCT (56)

where Favm,nct is the arithmetic mean air-vapor flowrate through the fill in each cooling
tower,

Favin ,nct  Favout ,nct

Favav ,nct  ; nct  NCT (57)
2
Optimal Design of Cooling Water Systems 171

and av ,nct is the harmonic mean density of the moist air through the fill calculated as:

 av ,nct  1  1 in ,nct  1 out ,nct  , nct  NCT (58)

cme
m e=1 e=2 e=3
(splash fill) (trickle fill) (film fill)
1 3.179688 7.047319 3.897830
2 1.083916 0.812454 0.777271
3 -1.965418 -1.143846 -2.114727
4 0.639088 2.677231 15.327472
5 0.684936 0.294827 0.215975
6 0.642767 1.018498 0.079696
Table 2. Constants for loss coefficients
The air-vapor flow at the fill inlet and outlet Favin,nct and Favout,nct are calculated as follows:

Favin ,nct  Fanct  win ,nct Fanct , nct  NCT (59)

Favout ,nct  Fanct  wout ,nct Fanct , nct  NCT (60)

where win,nct is the humidity (mass fraction) of the inlet air, and wout,nct is the humidity of the
outlet air. The required power for the cooling tower fan is given by:

Favin , nct Pt ,nct

PC f ,nct  ; nct  NCT (61)
in , nct  f ,nct

where  f ,nct is the fan efficiency. The power consumption for the water pump may be
expressed as (Leeper, 1981):


 g   FCUin L fi ,t  3.048
PC p    
  (62)
 gc   p 

where p is the pump efficiency. As can be seen in the equation (62), the power
consumption for the water pump depends on the total fill height (Lfi,t), which depends on
the arrangement of the cooling tower network (i.e., parallel (Lfi,t,pl) or series (Lfi,t,s));

L fi ,t  L fi ,t , pl  L fi ,t ,s (63)

If the arrangement is in parallel, the total fill height is equal to the fill height of the tallest
cooling tower, but if the arrangement is in series, the total fill height is the sum of the
cooling towers used in the cooling tower network. This decision can be represented by the
next disjunction,
172 Energy Management Systems

 3
znct , nct

    z3 
max nct , nct
FTTnct ,nct  FTTnct ,nct    
 min
 FTTnct ,nct  0 
FTTnct ,nct  FTTnct ,nct 

This last disjunction determines the existence of flowrates between cooling towers.
Following disjunction is used to activate the arrangement in series,

 zs4 
 
  znct 3 min
, nct    znct 3    zs4


 nct nct nct nct
,nct
  
     znct3
 0
 L fi ,t ,s   L fi ,s ,nct   nct nct 
, nct

 nctNCT   L 
 min   fi ,t , s  0 
 1
  znct 3
,nct 
 nct nct 
here  min z3 is the minimum number of interconnections between cooling towers when a
 nct ,nct
nct nct

series arrangement is used. The reformulation for this disjunction is the following:

 znct
3,1 min
, nct   Z nct ,nct
zs4,1 (64)
nct nct

L fi ,nct ,s  Lmax
fi
zs4 (65)

L fi ,t , s   L fi ,nct ,s (66)
nctNCT

If a series arrangement does not exist, then a parallel arrangement is used. In this case, the
total fill height is calculated using the next disjunction based on the Boolean variable Zp5,nct ,
which shows all possible combination to select the biggest fill height from the total possible
cooling towers that can be used in the cooling tower network:

 5,1
Zpl   5,2
Zpl   Zpl 5, LCT 
     
 L fi ,nct  1, pl  L fi , nct  1, pl   L fi ,nct  2 , pl  L fi ,nct  2 , pl  ....  L fi ,nct  LCT , pl  L fi ,nct  LCT , pl 
     
 L fi ,t , pl  L fi ,nct  1, p   L fi ,t , pl  L fi , nct  2 , p   L fi ,t , pl  L fi ,nct  LCT , pl 
The reformulation for the disjunction is:

5,1
zpl 5,2
 zpl 5, LCT
 ...  zpl  1  zs4   (67)

Notice that when zs4 is activated, then any binary variable zpl 5,nct
can be activated, but if zs4
5,nct
is not activated, only one binary variable zpl must be activated, and it must represent the
tallest fill. The rest of the reformulation is:

L fi ,nct  1, pl  L1fi ,nct  1, pl  L2fi , nct  1, pl  ...  LLCT

fi , nct  1, pl (68)
Optimal Design of Cooling Water Systems 173

L fi ,nct  2, pl  L1fi ,nct  2 , pl  L2fi , nct  2 , pl  ...  LLCT

fi , nct  2, pl (69)

L fi ,nct  LCT , pl  L1fi ,nct  LCT , pl  L2fi ,nct  LCT , pl  ...  LLCT
fi , nct  LCT , pl
(70)

L fi ,t , pl  L1fi ,t , pl  L2fi ,t , pl  ...  LLCT

fi ,t , pl (71)

L1fi ,nct  1, pl  L1fi ,nct  1, pl

L2fi , nct  2, pl  L2fi ,nct  2, pl
(72)

LLCT LCT
fi , nct  LCT , p  L fi , nct  LCT , pl

L1fi ,t , pl  L1fi ,nct  1, pl

L2fi ,t , pl  L2fi ,nct  2 , pl
(73)

LLCT LCT
fi ,t , pl  L fi , nct  NCT , pl

L1fi , nct  1, pl  max 5,1

L fi ,nct z pl

L2fi ,nct  1, pl  max 5,2

L fi ,nct zpl
(74)

LLCT max 5, LCT
fi , nct  1, pl  L fi ,nct zpl

L1fi ,nct  2, pl  max 5,1

L fi ,nct z pl

L2fi ,nct  2, pl  Lmax z5,2

fi ,nct pl
(75)

LNCT max 5, LCT
fi , nct  2, pl   L fi ,nct z pl

L1fi ,nct  LCT , pl  max 5,1

L fi ,nct z pl

L2fi ,nct  LCT , p  max 5,2

L fi ,nct zpl
(76)

LLCT max 5, LCT
fi , nct  LCT , pl   L fi ,nct z pl

Finally, an additional equation is necessary to specify the fill height of each cooling tower
depending of the type of arrangement,

L fi ,nct  L fi ,nct , pl  L fi ,nct ,s , nct  NCT (77)

174 Energy Management Systems

According to the thermodynamic, the outlet water temperature in the cooling tower must be
lower than the lowest outlet process stream of the cooling network and greater than the inlet
wet bulb temperature; and the inlet water temperature in the cooling tower must be lower
than the hottest inlet process stream in the cooling network. Additionally, to avoid the
fouling of the pipes, 50ºC usually are specified as the maximum limit for the inlet water
temperature to the cooling tower (Serna-González et al., 2010),

Twout ,nct  TWBin ,nct  2.8, nct  NCT (78)

Twout ,nct  TMPO  TMIN , nct  NTC (79)

Twin ,nct  TMPI  TMIN , nct  NTC (80)

Twin , nct  50º C , nct  NTC (81)

here TMPO is the inlet temperature of the coldest hot process streams, TMPI is the inlet
temperature of the hottest hot process stream. The final set of temperature feasibility
constraints arises from the fact that the water stream must be cooled and the air stream
heated in the cooling towers,

Twin ,nct  Twout ,nct , nct  NTC (82)

TAout ,nct  TAin ,nct , nct  NTC (83)

The local driving force (hsanct-hanct) must satisfy the following condition at any point in the
cooling tower (Serna-González et al., 2010),

hsan ,nct  han ,nct  0 n  1,..., 4; nct  NTC (84)

The maximum and minimum water and air loads in the cooling tower are determined by
the range of test data used to develop the correlations for the loss and overall mass transfer
coefficients for the fills. The constraints are (Kloppers and Kröger, 2003, 2005),

2.90  Fwin ,nct A fr ,nct  5.96, nct  NTC (85)

1.20  Fanct A fr ,nct  4.25, nct  NTC (86)

Although a cooling tower can be designed to operate at any feasible Fwin,nct/Fanct ratio,
Singham (1983) suggests the following limits:

0.5  Fwin ,nct Fanct  2.5, nct  NTC (87)

The flowrates of the streams leaving the splitters and the water flowrate to the cooling tower
have the following limits:

0  Fw1, j ,nct  Fw j , j  NEF ; nct  NCT (88)

Optimal Design of Cooling Water Systems 175

0  Fw2, j  Fw j j  NEF (89)

The objective function is to minimize the total annual cost of cooling systems (TACS) that
consists in the total annual cost of cooling network (TACNC), the total annual cost of cooling
towers (TACTC) and the pumping cost (PWC),

TACS  TACNC  TACTC  PWC (90)

PWC  HY cePC p (91)

where HY is the yearly operating time and ce is the unitary cost of electricity. The total
annual cost for the cooling network is formed by the annualized capital cost of heat
exchangers (CAPCNC) and the cooling medium cost (OPCNC).

TACNC  CAPCNC  OPCNC (92)

where the capital cooling network cost is obtained from the following expression,

 
CAPCNC  K F    CFHEi zi , k    CAHEi Ai, k  (93)
 iHP kST iHP kST 
Here CFHEi is the fixed cost for the heat exchanger i, CAHEi is the cost coefficient for the
area of heat exchanger i, KF is the annualization factor, and  is the exponent for the capital
cost function. The area for each match is calculated as follows,

Ai , k  qi , k Ui TMLi , k    (94)

U i  1  1 hi  1 hcu  (95)

where Ui is the overall heat-transfer coefficient, hi and hcu are the film heat transfer
coefficients for hot process streams and cooling medium, respectively. TMLi , k is the mean
logarithmic temperature difference in each match and  is a small parameter (i.e., 1x10 6 )
used to avoid divisions by zero. The Chen (1987) approximation is used to estimate
TMLi , k ,

 
1/3
TMLi , k   dtcali , k  dtfrii , k   dtcali , k  dtfrii , k  2  (96)
 
In addition, the operational cost for the cooling network is generated by the makeup
flowrate used to replace the lost of water in the cooling towers network,

OPCNC  CUwHY Fwr (97)

where CUw is the unitary cost for the cooling medium. The total annual cost of cooling
towers network involves the investment cost for the cooling towers (CAPTNC) as well as the
operational cost (OPTNC) by the air fan power of the cooling towers. The investment cost
for the cooling towers is represented by a nonlinear fixed charge expression of the form
(Kintner-Meyer and Emery, 1995):
176 Energy Management Systems

CAPTNC  KF  CCTF znct


2
 CCTVnct A fr ,nct L fi , nct  CCTMA Fanct  (98)
nctNCT

where CCTF is the fixed charge associated with the cooling towers, CCTVnet is the incremental
cooling towers cost based on the tower fill volume, and CCTMA is the incremental cooling
towers cost based on air mass flowrate. The cost coefficient CCTVnet depends on the type of
e
packing. To implement the discrete choice for the type of packing, the Boolean variable Ynct
is used as part of the following disjunction,

 1
Ynct   2
Ynct   3
Ynct 
     
  splash fill     trickle fill      film fill  
 1   2   3 
CCTV nct  CCTV  
nct  CCTV nct  CCTV  
nct  CCTV nct  CCTV 
nct 

This disjunction is algebraically reformulated as:

1 2 3
CCTVnct  CCTVnct  CCTVnct  CCTVnct , nct  NCT (99)

e
CCTVnct  a e ynct
e
, e  1,..., 3, nct  NCT (100)

where the parameters ae are 2,006.6, 1,812.25 and 1,606.15 for the splash, trickle, and film
types of fill, respectively. Note that the investment cost expression properly reflects the
influence of the type of packing, the air mass flowrate (Fanet) and basic geometric
parameters, such as height (Lfi,nct) and area (Afi,nct) for each tower packing. The electricity cost
needed to operate the air fan and the water pump of the cooling tower is calculated using
the following expression:

OPTNC  HY ce  PC f ,nct (101)

nct  1

This section shows the physical properties that appear in the proposed model, and the
property correlations used are the following. For the enthalpy of the air entering the tower
(Serna-González et al., 2010):

hain  6.4  0.86582 * TWBin  15.7154 exp  0.0544 * TWBin  (102)

For the enthalpy of saturated air-water vapor mixtures (Serna-González et al., 2010):

hsai  6.3889  0.86582 * Twi  15.7154 exp  0.054398 * Twi  , i  1,...4 (103)

For the mass-fraction humidity of the air stream at the tower inlet (Kröger, 2004):

 2501.6  2.3263 TWBin    0.62509  PVWB, in  

win    
 2501.6  1.8577 TA   4.184 TWB    Pt  1.005  PVWB ,in  
 in in   
(104)
 1.00416 TAin  TWBin  
 
 2501.6  1.8577  TAin   4.184  TWB 
in 
Optimal Design of Cooling Water Systems 177

where PVWB,in is calculated from Equation (115) and evaluated at T = TWBin. For the mass-
fraction humidity of the saturated air stream at the cooling tower exit (Kröger, 2004):

0.62509 PVout
wout  (105)
Pt  1.005 PVout

where PVout is the vapor pressure estimated with Equation (115) evaluated at T = TAout, and
Pt is the total pressure in Pa. Equation (115) was proposed by Hyland and Wexler (1983) and
is valid in the range of temperature of 273.15 K to 473.15 K,

3
ln  PV    cnT n  6.5459673 ln T  (106)
n 1

PV is the vapor pressure in Pa, T is the absolute temperature in Kelvin, and the constants
have the following values: c-1 = 5.8002206 x 103, c0 = 1.3914993, c1 = -4.8640239 x 10-3, c2 =
4.1764768 x 10-5 and c3 = -1.4452093 x 10-7. For the outlet air temperature, Serna-González et
al. (2010) proposed:

hsaout  6.38887667  0.86581791 * TAout  15.7153617 exp  0.05439778 * TAout   0 (107)

For the density of the air-water mixture (Serna-González et al., 2010):

Pt  w 
 1- 1  w (108)
287.08 T  w  0.62198 


where Pt and T are expressed in Pa and K, respectively. The density of the inlet and outlet
air are calculated from the last equation evaluated in T = TAin and T = TAout for w = win and
w = wout, respectively.

3. Results
Two examples are used to show the application of the proposed model. The first example
involves three hot process streams and the second example involves five hot process
streams. The data of these examples are presented in Table 3. In addition, the value of
parameters ce, HY, KF, NCYCLES, ηf, ηp ,Pt, CCTF, CCTMA, CUw, CPcu, β, CFHE, CAHE are 0.076
$US/kWh, 8000 hr/year, 0.2983 year-1, 4, 0.75, 0.6, 101325 Pa, 31185 $US, 1097.5 $US/(kg dry
air/s), 1.5449x10-5 $US/kg water, 4.193 kJ/kg°C, 1, 1000$US, 700$US/m2, respectively. For
the Example 1, fresh water at 10 °C is available, while the fresh water is at 15°C for the
Example 2.
For the Example 1, the optimal configuration given in Figure 3 shows a parallel arrangement
for the cooling water network. Notice that one exchanger for each hot process stream is
required. In addition, only one cooling tower was selected; consequently, the cooling tower
network has a centralized system for cooling the hot process streams. The selected packing
is the film type, and the lost water is 13.35 kg/s due to the evaporation lost (75%), and the
drift and blowdown water (4.89% and 20.11%), while a 70.35% of the total power
consumption is used by the fan and the rest is used by the pump (29.64%). The two above
terms represent the total operation cost of the cooling system; therefore, both the evaporated
water and the power fan are the main components for the cost in this example. Notice that
178 Energy Management Systems

the water flowrate in the cooling network is 326.508 kg/s, but the reposition water only is
13.25 kg/s, which represents a save of freshwater of 95.94% respect to the case when is not
used a cooling tower for thermal treatment of the cooling medium. The total annual cost is
468,719.906$US/year. The contribution to total annual cost for the cost of cooling network is
66%, while for cooling tower network and the pump are 31% and 2.96%, respectively. These
results are given in the Table 4.
Respect to the Example 2, Figure 4 presents the optimal configuration, which shows a
parallel arrangement to the cooling water network, while the cooling towers network is
formed by a distributed system composed by two cooling towers to treat the effluents from
the cooling network and to meet the cooling requirements. The selected fill is the film type,
and the lost water by evaporation, drift and blowdown represent a 74.99%, 3.94% and
21.07% of the total water lost, respectively. Respect the total power consumption in the
cooling system, the fan demands a 65.37% and the pump use a 34.62% of the total cost. The
economical results are given in the Table 4. The optimal cooling system shows costs for the
cooling network, cooling tower and water pump equal to 61.21%, 36.34% and 2.44%,
respectively, of the total annual cost. In addition, for the case that only one cooling tower is
selected, the total annual cost is 143,4326.66$US/year, which is 7% more expensive than the
optimal configuration. The savings obtained are because the distributed system is able to
find a better relationship between the capital cost and the operation cost, which depends of
the range, inlet water flowrate and inlet air flowrate to the cooling tower network; therefore,

Fig. 3. Optimal configuration for the Example 1

Optimal Design of Cooling Water Systems 179

in the distributed systems there are more options. In this case, the use of freshwater by the
cooling network is reduced by 94.92% with the use of the cooling towers. Other advantage
of use a distributed system is that depending of the problem data just one cooling tower
could not meet with the operational and/or thermodynamic constraints and could be
necessary to use more than one cooling tower.

Example 1
Streams THIN (ºC) THOUT (ºC) FCP (kW/ºC) Q (kW) h (kW/m2ºC)
1 40 76.6 100 3660 1.089
2 60 82 60 1320 0.845
3 45 108.5 400 25540 0.903
Example 2
Streams THIN (ºC) THOUT (ºC) FCP (kW/ºC) Q (kW) h (kW/m2ºC)
1 80 60 500 10000 1.089
2 75 28 100 4700 0.845
3 120 40 450 36000 0.903
4 90 45 300 13500 1.025
5 110 40 250 17500 0.75
Table 3. Data for examples

Example 1 Example 2
TACS
468,719.906 1,334,977.470
(US$/year)
TACNC
309,507.229 817,192.890
(US$/year)
TACTC
145,336.898 485,196.940
(US$/year)
OPCNC
6,131.013 16,588.250
(US$/year)
OPTNC
32,958.635 61,598.140
(US$/year)
PWC
13,875.780 32,587.640
(US$/year)
CAPCNC
303,376.216 800,604.640
(US$/year)
CAPTNC
112,378.262 181,000.330
(US$/year)
Table 4. Results for examples
180 Energy Management Systems

Fig. 4. Optimal configuration for the Example 2

4. Conclusion
This chapter presents a new model for the detailed optimal design of re-circulating cooling
water systems. The proposed formulation gives the system configuration with the minimum
total annual cost. The model is based on a superstructure that considers simultaneously
series and parallel arrangements for the cooling water network and cooling tower network,
in which the cooling medium can be thermally treated using a distributed system.
Significant savings were obtained with the distributed cooling systems for the
Optimal Design of Cooling Water Systems 181

interconnection between cooling water network and cooling towers. Evaporation represents
the main component for the lost of water (70-75%); while the drift and blowdown represent
the 3-5% and 20-25%, respectively. The fan power consumption usually represents the 65-
70% of the total power consumption in the cooling system; and the pump represents around
the 30-35%. For re-circulating cooling water systems the costs of cooling network, cooling
tower network and the water pump represent the 60-70%, 30-40% and 2-5% of the total
cooling system cost, respectively. When re-circulating cooling water systems are used, the
use of freshwater in the cooling network is significantly reduced (i.e., 95%).

5. References
Chen, J.J. (1987). Letter to the Editors: Comments on improvement on a replacement for the
logarithmic mean. Chemical Engineering Science, Vol.42, pp. 2488-2489.
Feng, X.; Shen, R.J. & Wang, B. (2005). Recirculating cooling-water network with an
intermediate cooling-water main. Energy & Fuels, Vol.19, No.4, pp. 1723-1728.
Hyland, R.W. & Wexler, A. (1983). Formulation for the thermodynamic properties of the
saturated phases of H2O from 173.15 K and 473.15 K. ASHRAE Transactions,
Vol.89,No.2A, pp. 500-519.
Kemmer, F.N. (1988). The NALCO Water Handbook, McGraw-Hill, New York, USA.
Kim, J.K. & Smith, R. (2001). Cooling water system design. Chemical Engineering Science,
Vol.56, No.12, pp. 3641-3658.
Kintner-Meyer, M. & Emery, A.F. (1995). Cost-optimal design for cooling towers. ASHRAE
Journal, Vol.37, No.4, pp. 46-55.
Kloppers, J.C. & Kröger, D.G. (2003). Loss coefficient correlation for wet-cooling tower fills,
Applied Thermal Engineering, Vol.23, No.17, pp. 2201-2211.
Kloppers, J.C. & Kröger, D.G. (2005). Refinement of the transfer characteristic correlation of
wet-cooling tower fills. Heat Transfer Engineering, Vol.26, No.4, pp. 35-41.
Kröger, D.G. (2004). Air-Cooled Heat Exchangers and Cooling Towers, PenWell Corp., Tulsa,
USA.
Leeper, S.A. (1981). Wet cooling towers: rule-of-thumb design and simulation. Report, U.S.
Department of Energy, 1981.
Merkel, F. (1926).Verdunstungskuhlung, VDI Zeitchriff Deustscher Ingenieure, Vol.70, pp. 123-
128.
Mohiudding, A.K.M. & Kant, K. (1996). Knowledge base for the systematic design of wet
cooling towers. Part I: Selection and tower characteristics. International Journal of
Refrigeration, Vol.19, No.1, pp. 43-51.
Ponce-Ortega, J.M.; Serna-González, M. & Jiménez-Gutiérrez, A. (2007). MINLP synthesis of
optimal cooling networks. Chemical Engineering Science, Vol.62, No.21, pp. 5728-
5735.
Serna-González, M.; Ponce-Ortega, J.M. & Jiménez-Gutiérrez, A. (2010). MINLP
optimization of mechanical draft counter flow wet-cooling towers. Chemical
Engineering Research and Design, Vol.88, No.5-6, pp. 614-625.
Singham, J.R. (1983). Heat Exchanger Design Handbook, Hemisphere Publishing Corporation,
New York, USA.
Söylemez, M.S. (2001). On the optimum sizing of cooling towers. Energy Conversion and
Management, Vol.42, No.7, pp. 783-789.
 182 Energy Management Systems

Söylemez, M.S. (2004). On the optimum performance of forced draft counter flow cooling
towers. Energy Conversion and Management, Vol.45, No.15-16, pp. 2335-2341.
Vicchietti, A., Lee, S. & Grossmann, I.E. (2003). Modeling of discrete/continuous
optimization problems characterization and formulation of disjunctions and their
relaxations. Computers and Chemical Engineering, Vol.27, No.3, pp. 433-448.

View publication stats