A new approach for optimization of thermal power plant based

on the exergoeconomic analysis and structural optimization

method: Application to the CGAM problem
Seyyed Masoud Seyyedi
*
, Hossein Ajam, Said Farahat
Department of Mechanical Engineering, University of Sistan and Baluchestan, Zahedan 98164, Iran
a r t i c l e i n f o
Article history:
Received 28 July 2009
Accepted 21 March 2010
Available online 24 April 2010
Keywords:
Optimization
Exergoeconomic analysis
Structural method
a b s t r a c t
In large thermal systems, which have many design variables, conventional mathematical optimization
methods are not efcient. Thus, exergoeconomic analysis can be used to assist optimization in these
systems. In this paper a new iterative approach for optimization of large thermal systems is suggested.
The proposed methodology uses exergoeconomic analysis, sensitivity analysis, and structural optimiza-
tion method which are applied to determine sum of the investment and exergy destruction cost ow
rates for each component, the importance of each decision variable and minimization of the total cost
ow rate, respectively. Applicability to the large real complex thermal systems and rapid convergency
are characteristics of this new iterative methodology. The proposed methodology is applied to the bench-
mark CGAM cogeneration system to show how it minimizes the total cost ow rate of operation for the
installation. Results are compared with original CGAM problem.
2010 Elsevier Ltd. All rights reserved.
1. Introduction
The development of design techniques for an energy system
with minimized costs is a necessity in a world with nite natural
resources and the increase of the energy demand in developing
countries [1]. Optimization has always been one of the most inter-
ested and essential subjective in the design of energy systems.
Usually we are interested to know optimum conditions of thermal
systems. Thus we need methods for optimization of such systems.
In large complex thermal systems, which have many design vari-
ables, conventional mathematical optimization methods are not
efcient. Thus, exergoeconomic analysis can be used to assist opti-
mization in these systems. On the other hand, complex thermal
systems cannot always be optimized using mathematical optimi-
zation techniques. The reasons include incomplete models, system
complexity and structural changes [2].
Exergoeconomic (Thermoeconomic) is the branch of engineer-
ing that combines exergy analysis with economic constraints to
provide the system designer with information not available
through conventional energy analysis and economic evaluation
[3]. The objective of a thermoeconomic analysis might be: (a) to
calculate separately the cost of each product generated by a system
having more than one product; (b) to understand the cost forma-
tion process and the ow of costs in the system; (c) to optimize
specic variables in a single component; or (d) to optimize the
overall system [2]. A thermodynamic optimization aims at mini-
mizing the thermodynamic inefciencies: exergy destruction and
exergy loss. The objective of a thermoeconomic optimization, how-
ever, is to minimize costs, including costs owing to thermodynamic
inefciencies [4].
In 1994, a cogeneration plant, known as the CGAM problem,
was dened as a test case by a group of concerned specialists in
the led of exergoeconomic, in order to compare their different
thermoeconomic methodologies [59]. Exergoeconomic methods
can be grouped in two classes: the algebraic methods and the cal-
culus methods [10,11]. All of these methods are based on an exer-
goeconomic model, which basically consists of an interposed set of
linear exergy equations that dene the productive objective of
each component of the plant [3]. Some of the algebraic methods
are: exergetic cost theory (ECT) [12], average cost theory (ACT)
[4], specic cost exergy costing method (SPECO) [13] and modied
productive structural analysis (MOPSA) [14,15]. Furthermore,
some of the calculus methods are: thermoeconomical functional
analysis (TFA) [16,17] and engineering functional analysis (EFA)
[18]. Then, in 1992, Erlach et al. [19] developed a common mathe-
matical language for exergoeconomics, called the structural theory
of thermoeconomics. Furthermore, Hua et al. [20], El-Sayed [21],
Benelmir and Feidt [22] have proposed decomposition strategies
based on second law reasoning to reduce complexity in the
optimization of complete systems. A critical review of relevant
publications regarding exergy and exergoeconomic analysis can
be found in articles by Leonardo et al. [23], Sahoo [3] and Zhang
et al. [24]. In 1997, Tsatsaronis and Moran [2], showed how certain
0196-8904/$ - see front matter 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.enconman.2010.03.014
* Corresponding author. Tel.: +98 541 2426206; fax: +98 541 2447092.
E-mail address: s.masoud_seyedi@yahoo.com (S.M. Seyyedi).
Energy Conversion and Management 51 (2010) 22022211
Contents lists available at ScienceDirect
Energy Conversion and Management
j our nal homepage: www. el sevi er . com/ l ocat e/ enconman
exergy-related variables can be used to minimize the cost of a ther-
mal system. They applied this iterative optimization technique to
the benchmark CGAM problem. In 2004, Leonardo et al. [23] pre-
sented the development and automated implementation of an iter-
ative methodology for exergoeconomic improvement of thermal
systems integrated with a process simulator, so as to be applicable
to real, complex plants. Also, see Refs. [25,26]. Most exergoeco-
nomic optimization theories have been applied to relatively simple
systems only. Conventional mathematical optimization, exergo-
economic or not, of real thermal systems are large scale problems,
due to their complicated nonlinear characteristics and because the
mass, energy and exergy (or entropy) balance equations must be
introduced in the problem as restrictions [23].
In this paper, a new iterative method for the optimization of
thermal systems is developed using exergoeconomic analysis, sen-
sitivity analysis, and structural optimization method. Exergo-
economoic analysis is used to determine sum of the investment
and exergy destruction cost ow rates for each component. A
numerical sensitivity analysis is performed in order to determine
the importance of each decision variable. Finally, the total cost ow
rate is minimized and the optimum vector of decision variables is
determined by using structural optimization method. The advanta-
ges of this new iterative method are: (1) it can be applied to the
real complex large thermal systems; (2) the procedure of optimiza-
tion is performed without user interface, i.e. there is no to the deci-
sion of designer in each iteration, and (3) since it uses a numerical
sensitivity analysis, convergency is improved. In order to represent
how this new methodology can be used for optimization of real
complex large thermal systems, it is applied to the benchmark
CGAM cogeneration system as a test case and results are compared
with the original CGAM problem.
2. CGAM problem
In 1990, a group of concerned specialists in the led of exergo-
economic (C. Frangopoulos, G. Tsatsaronis, A. Valero, and M. von
Spakovsky) decided to compare their methodologies by solving a
predened and simple problem of optimization: the CGAM prob-
lem, which was named after the rst initials of the participating
investigators. The objective of the CGAM problem was to show
how the methodologies were applied, what concepts were used
and what numbers were obtained in a simple and specic problem.
In the nal analysis, the aim of CGAM problem was the unication
of exergoeconomic methodologies [3]. The CGAM system refers to
a cogeneration plant which delivers 30 MW of electricity and
14 kg s
1
of saturated steam at 20 bar. A schematic of cogeneration
plant is shown in Fig. 1. The system consists of an air compressor
(AC), an air preheater (APH), a combustion chamber (CC), a gas-tur-
bine (GT) and a heat-recovery steam generator (HRSG). The envi-
ronment conditions are dened as T
0
= 298.15 K and P
0
= 1.013 bar.
The objective function is the total cost ow rate of operation for
the installation that is obtained from
_
C
T
_ m
f
c
f
LHV

5
i1
_
Z
i
1
where C

T
($/s) is the total cost ow rate of fuel and equipment and
_
Z
i
in ($/s) is the cost ow rate associated with capital investment
and the maintenance cost for the ith component (i = AC, CC, GT,
APH, HRSG).
Also, exergetic efciency of the cycle (g
II
) is dened as:
g
II

_
W
net
_ m
s
e
9
e
8

_ m
f
e
f
2
The key design variables, (the decision variables), for the cogen-
eration system are the compressor pressure ratio PR, the isentropic
compressor efciency g
AC
, the isentropic turbine efciency g
GT
, the
temperature of the air entering the combustion chamber T
3
, and
the temperature of the combustion products entering the gas tur-
bine T
4
. The objective is to minimize Eq. (1) subject to the con-
straints imposed by the physical, thermodynamic and cost
models of the installation. For more details see Appendix A and
Ref. [5].
Nomenclature
c cost per exergy unit ($/kJ)
C

exergetic cost ow rates ($/s)

com component
CRF capital recovery factor
E

exergy ow rate (kW)

i ith plant component
_
I irreversibility rate (kW)
j jth decision variable
LHV lower heating value of fuel (kJ/kg)
_ m mass ow rate
N number of the hours of plant operation per year (h/year)
p parameter for sensitivity analysis, expressions (5) and
(6)
PR pressure ratio
T temperature (K)
_
W
net
net work of the cycle (kW)
x decision variable
X vector of decision variables
Z purchase costs of the ith component ($),
_
Z investment cost ow rate ($/s)
Greek letters
a user prescribed tolerance for the iterative process, Eq.
(7)
e component exergetic efciencies
f capital cost coefcient
g isentropic efciency
g
II
exergetic efciency of the cycle
l dened in Eq. (8)
r coefcient of structural bonds
u maintenance factor
Subscripts
0 index for environment (reference state)
AC air compressor
APH air preHeater
CC combustion chamber
D destruction
f fuel
GT gas-turbine
HRSG heat-recovery steam generator
In inlet
Iter iteration
k kth plant component
L lower
OPT optimum
Out outlet
P product
S steam
T total
U upper
S.M. Seyyedi et al. / Energy Conversion and Management 51 (2010) 22022211 2203
3. Structural optimization method
The purpose of this optimization is to determine the capital cost
of a selected component (system element) corresponding to the
minimum annual operating cost of the plant with a given plant
output and thus, by implication, corresponding to the minimum
unit cost of the product [27]. The following relation for the kth
component must be satised until the total cost ow rate of oper-
ation for the installation, C

T
, be minimized:
@
_
I
k
@x
i

OPT

1
c
I
k;i
@
_
Z
k
@x
i

3
In the proposed methodology, Eq. (3) is numerically calculated
using Newtons nite difference formula:
_
I
k
x
i
1
_
I
k
x
i

Dx
i


1
c
I
k;i
_
Z
k
x
i1

_
Z
k
x
i

Dx
i

4
where Dx
i
x
i
1 x
i
. For more details see Appendix B and Ref.
[27].
4. Proposed iterative methodology
In this paper, we have proposed a new approach for optimiza-
tion of complex thermal power plant. It is worth to mention that
structural optimization method (see Section 3 and Appendix B) is
a method of optimization for one component of the system, so that
the total cost ow rate of the system to be minimized while we
have applied it for optimization of complex thermal power plant
using exergoeconomic analysis (here, ACT method [4]) and numer-
ical sensitivity analysis.
4.1. Numerical sensitivity analysis
A numerical sensitivity analysis is performed for each design
variable (decision variable) in order to determine the importance
of each decision variable. The sensitivity analysis is performed
for each x
i
according to the following procedure:
if
x
i
Dx
i
Dg
II
g
II
> p; x
i
affects exergetic efficiency; and 5
if
x
i
Dx
i
D
_
C
T
_
C
T
> p; x
i
affects total cost flow rate of the system
6
Thus, each decision variable will be settled in one of the follow-
ing groups:
Group 1: if decision variable affects both C

T
and g
II
.
Group 2: if decision variable affects on the C

T
only.
Group 3: if decision variable affects on the g
II
only.
Group 4: if decision variable affects neither C

T
nor g
II
.
The left-hand sides of expressions (5) and (6) are numerically
evaluated and compared to the value of the parameter p where
p 0:1x
i
=Dx
i
.
4.2. Algorithm for the proposed iterative methodology
Fig. 2 represents a general ow diagram of the proposed itera-
tive methodology. More details of the general ow diagram are
presented as follows:
Step 1. Selecting vector X = [x
1
, x
2
, . . . , x
n
] where n is the number
of decision variables. Then, determining the lower limit
value (x
i,L
) and the upper limit value (x
i,U
) for each decision
variable (x
i
).
Step 2. Thermodynamic analysing and checking for whether X is a
feasible solution or not. If X is a feasible solution, iter = 1. If
X is infeasible, go to step 1.
Step 3. Exergoeconomic analysing using ACT method [4].
Step 4. Sorting components with decreasing order in terms of the
sum of the investment and exergy destruction cost ow
rates (C

D,k
+
_
Z
D,k
). Setting F
CZ
= [com
1
, com
2
, . . . , com
m
]
where m is the number of components and F
CZ
is a vector
that includes names of components corresponding to
(C

D,k
+
_
Z
D,k
) in decreasing order.
Step 5. Numerical sensitivity analysing and classifying each x
j
of
the vector X as:
X

1
; x

2
; :::; x

g
....
Group1
; x

g1
; x

g2
; . . . ; x

l
....
Group2
; x

l1
; x

l2
; . . . x

k
....
Group3
; x

k1
; x

k2
; . . . x

n
....
Group4

where x

j
is the same as x
j
that set in groups 1, 2, 3 or 4.
Step 6. Performing structural optimization method:
for i = 1 to m
for j = 1 to n
Fig. 1. Schematic of CGAM problem.
2204 S.M. Seyyedi et al. / Energy Conversion and Management 51 (2010) 22022211
6.i.j. For the ith element of F
CZ
(com
i
) and the jth element
of vector X

j
, structural optimization is implemented. Thus,
x

j
is updated and replaced by this value in vector X

.
end
Vector X

is updated which is called X

com,i
.
end
Step 7. Checking for convergency:
7.1. X

is named X
iter
and total cost ow rate (C

T
) is calculated
and named C

T,iter
.
7.2. The nal vector of X (here, X
com,m
) is named X
iter+1
and total
cost ow rate (C

T
) is calculated and named C

T,iter+1
.
7.3. Checking the following inequality:
_
C
T; iter1

_
C
T; iter

_
C
T; iter
< a 7
If Eq. (7) is satised, X
iter+1
is the optimum solution otherwise, X is
replaced by X
iter+1
and the procedure is repeated from step 3 and set
iter = iter + 1.
In Eq. (7), a is a small positive value.
It should be noted that step 5 is preformed in order to assist ra-
pid convergency.
4.3. Application to the CGAM problem
In order to represent how this proposed methodology optimizes
thermal power plant, CGAM problem is selected as a test case. All
codes for calculations were developed in MATLAB. For each deci-
sion variable x
i
, the lower (x
i,L
) and the upper (x
i,U
) limiting values
should be determined with predened steps that have been pre-
sented in Table 1. The following steps are performed respectively:
Step 1. A vector of decision variables is randomly selected.
X PR; T
3
; T
4
; g
AC
; g
GT

For each decision variable (x

i
), the lower limit value (x
i,L
) and the
upper limit value (x
i,U
) should be determined.
Step 2. A Thermodynamic analysis is performed to determine
whether X is a feasible solution or not. If X is a feasible
solution iter = 1. If X is infeasible, go to step 1.
Start
Select X = [ x1, x2 , .xn]
Evaluate objective function, i.e. total cost flow rate (T) by X
Exergoeconomic analysis (ACT method)
& Evalue (D,k + D,k) for each component
Evaluate objective function, i.e. total cost flow rate ( T
) by Xnew
Numerical sensitivity analysis and classify each xj in vector X that
] ,... , , ,... , , ,..., , , ,..., , [
4
* *
2
*
1
3
* *
2
*
1
2
* *
2
*
1
1
* *
2
*
1
*
Group
n k k
Group
k l l
Group
L g g
Group
g
x x x x x x x x x x x x X
+ + + + + +
=
STOP
NO
Yes
Convergence
Criteria?
Components are ranked in decreasing order in terms of the sum of the investment and
exergy destruction cost flow rates (D,k + D,k), i.e. FCZ = [com1, com2,,comm]
Print optimum vector ( XOPT ) & optimum total cost flow rate T,OPT
For all components of Vector FCZ & all decision variables of
vector X
*
Perform Structural Optimization
X replaced by Xnew
Fig. 2. General ow diagram of the proposed iterative methodology.
Table 1
The lower and the upper limiting values and steps for the decision variables of the
CGAM cogeneration system.
Variable Value Step
Minimum Maximum
PR 5 25 0.01
T
3
(K) 500 1200 0.5
T
4
(K) 1200 1800 0.5
g
AC
0.7 0.9 0.0001
g
GT
0.7 0.92 0.0001
S.M. Seyyedi et al. / Energy Conversion and Management 51 (2010) 22022211 2205
Step 3. An exergoeconomic system analysis is performed to deter-
mine all exergy ow rates (E

i
) , exergy destruction ow
rates (E

D,k
), component exergetic efciencies (e
k
) , exerget-
ic cost ow rates (C

i
), component product (c
P,k
) and fuel
(c
F,k
) specic exergetic costs, exergy destruction cost ow
rates (C

D,k
), component investment cost ow rates (
_
Z
k
),
and system total ow cost (C

T
) . Detailed information of
how to calculate these variables can be found in the article
by Tsatsaronis [2] and in the book by Bejan et al. [4].
Step 4. In this step components are ranked in decreasing order in
terms of the sum of the investment and exergy destruction
cost ow rates (C

D,k
+
_
Z
D,k
). Assume, for example, that the
maximum values of (C

D,k
+
_
Z
D,k
) are corresponding to CC,
GT, HRSG, APH and AC, respectively. Thus, F
CZ
= [CC, GT,
HRSG, APH, AC].
Step 5. A numerical sensitivity analysis is performed to determine
the importance of decision variables. Thus, decision vari-
ables are grouped in four classes. Assume X

= [T
4
, T
3
, PR,
g
GT
, g
AC
].
Step 6. Performing structural optimization method for all compo-
nents as follows:
6.1. For the rst element of F
CZ
(here, CC)
6.1.1. For the rst element of vector X

(here, T
4
) structural
optimization is performed. Thus, T
4
is updated and
replaced by this value in vector X

.
6.1.2. For the second element of vector X

(here, T
3
) struc-
tural optimization is performed. Thus, T
3
is updated
and replaced by this value in vector X

.
6.1.3. For the third element of vector X

(here, PR) structural

optimization is performed. Thus, PR is updated and
replaced by this value in vector X

.
6.1.4. For the fourth element of vector X

(here, g
GT
) struc-
tural optimization is performed. Thus, g
GT
is updated
and replaced by this value in vector X

.
6.1.5. For the fth element of vector X

(here, g
AC
) structural
optimization is performed. Thus, g
AC
is updated and
replaced by this value in vector X

.
Thus, vector X

is updated and named X

CC
.
6.2. For the second element of F
CZ
(here, GT), the all above
operations (6.1.16.1.5) are repeated for elements of
vector X
CC
. For example for the rst element, 6.2.1 is
as follows:
6.2.1. For the rst element of vector X
CC
(here, T
4
) structural
optimization is performed. Thus, T
4
is updated and
replaced by this value in vector X
CC
.
Thus, after doing operations 6.2.16.2.5, vector X
CC
is up-
dated and named X
GT
.
6.3. For the third element of F
CZ
(here, HRSG), the all
above operations (6.1.16.1.5) are repeated for ele-
ments of vector X
GT
. For example for the rst ele-
ment, 6.3.1 is as follows:
6.3.1. For the rst element of vector X
GT
(here, T
4
) struc-
tural optimization is performed. Thus, T
4
is updated
and replaced by this value in vector X
GT
.
Thus, after doing operations 6.3.16.3.5, vector X
GT
is up-
dated and named X
HRSG
.
6.4. For the fourth element of F
CZ
(here, APH), the all above
operations (6.1.16.1.5) are repeated for elements of
vector X
HRSG
. For example for the rst element, 6.4.1
is as follows:
6.4.1. For the rst element of vector X
HRSG
(here, T
4
) struc-
tural optimization is performed. Thus, T
4
is updated
and replaced by this value in vector X
HRSG
.
Thus, after doing operations 6.4.16.4.5, vector X
HRSG
is up-
dated and named X
APH
.
6.5. For the fth element of F
CZ
(here, AC), the all above
operations (6.1.16.1.5) are repeated for elements of
vector X
APH
. For example for the rst element, 6.5.1
is as follows:
6.5.1. For the rst element of vector X
APH
(here, T
4
) structural
optimization is performed. Thus, T
4
is updated and
replaced by this value in vector X
APH
.
Thus, after doing operations 6.5.16.5.5, vector X
APH
is up-
dated and named X
AC
.
Step 7. Checking for convergency:
7.1. X

is named X
iter
and total cost ow rate (C

T
) is calcu-
lated and named C

T,iter
.
7.2. The nal vector of X (here, X
AC
) is named X
iter+1
and
total cost ow rate (C

T
) is calculated and named
C

T,iter+1
.
7.3. Checking Eq. (7).
5. Results and discussion
Table 2 represents a vector of decision variables that has been
randomly selected by the program. The table represents how the
selected vector is updated in each iteration. Also, in each iteration,
the value of the objective function (system total cost ow rate) has
been shown. The last row shows the optimum vector, i.e. the val-
ues of obtained decision variables that minimize objective func-
tion. Fig. 3 shows sum of the investment cost ow rate (
_
Z
T
), fuel
cost ow rate (C

F
) and total cost ow rate (C

T
) in each iteration cor-
responding to Table 2. The gure shows how sum of the
_
Z
T
and C

F
,
i.e. C

T
, is converged. Fig. 4 shows the total cost ow rate (C

T
) with
respect to PR, T
3
, T
4
and g
GT
corresponding to data of Table 2. Fig. 5
shows exergetic efciency (g
II
) with respect to the total cost ow
rate (C

T
) corresponding to Table 2. The gure shows values of exer-
getic efciency in each iteration with respect to the total cost ow
rate. It is noticeable that in optimumconditions, i.e. minimumtotal
cost ow rate, exergetic efciency is not maximum. Furthermore,
the gure shows that in the third iteration exergetic efciency is
maximum. Table 3 shows some vectors of decision variables that
have been randomly selected by the program. This table shows se-
lected initial vectors and corresponding nal (optimum) vectors for
Table 2
Steps to obtain the optimum solution.
Iteration X C

T
($/s)
PR T
3
(K) T
4
(K) g
AC
g
GT
Initial 12.5 815 1392 0.75 0.75 0.797919
1 9.76 863.5 1464 0.8554 0.8916 0.366426
2 9.54 873.5 1475.5 0.8552 0.8909 0.364635
3 9.12 887 1483.5 0.8520 0.8886 0.363199
4 9.14 896 1482.5 0.8474 0.8832 0.362588
5 9.05 899 1483.5 0.8469 0.8823 0.362438
6 8.92 902.5 1484 0.8475 0.8805 0.362333
7 8.88 907 1487 0.8476 0.8783 0.362237
8 8.64 909.5 1487 0.8473 0.8786 0.362149
9 8.62 911.5 1490.5 0.8470 0.8787 0.362052
10 8.62 914 1492 0.8470 0.8782 0.362039
11 8.52 914 1492 0.8470 0.8782 0.362033
2206 S.M. Seyyedi et al. / Energy Conversion and Management 51 (2010) 22022211
each case. In addition, the value of the objective function
(system total cost ow rate) in each case has been represented. l
in the last column shows the percent of relative error of the
optimum solution results from Ref. [5] and the total cost ow rate
(C

T
) from the present work that is calculated by the following
equation:
l
_
C
T

_
C
T;OPT
_
C
T;OPT

100 8
where
_
C
T;OPT
is the optimum solution from Ref. [5] (see Tables 4 and
5). Table 3 also shows that the number of iterations increases when
Fig. 3. Sum of the investment cost ow rate (
_
Z
T
), fuel cost ow rate (C

F
) and total cost ow rate (C

T
) with respect to iteration.
Fig. 4. Total cost ow rate (C

T
) with respect to (a) PR, (b) T
3
, (c) T
4
and (d) g
GT
.
S.M. Seyyedi et al. / Energy Conversion and Management 51 (2010) 22022211 2207
both T
3
and T
4
in the initial vector are more than those of real values
in the optimum solution (e.g. see cases 1 and 7). Also, the table
shows that the number of iterations decreases when T
3
in the initial
vector is near to its real value in the optimum conditions (e.g. see
Table 3
Some test cases to optimum solution.
Case X C

T
($/s) No. iteration l (%)
PR T
3
(K) T
4
(K) g
AC
g
GT
1 Initial 10 1000 1720 0.88 0.78 4.936762 1264
Final 8.37 920.5 1493.5 0.8447 0.8763 0.362089 24 0.020
2 Initial 18.5 860 1720 0.89 0.88 15.623946 4215
Final 8.52 912.5 1491 0.8464 0.8794 0.362033 22 0.003
3 Initial 15 850 1682 0.83 0.80 2.099394 480
Final 8.5 917.5 1492 0.8432 0.8777 0.362080 13 0.016
4 Initial 10 900 1550 0.88 0.83 0.411631 13.7
Final 8.64 914 1493 0.8456 0.8783 0.362037 6 0.004
5 Initial 8 680 1532 0.88 0.88 0.457613 26.4
Final 8.57 912 1493 0.8459 0.8803 0.362051 6 0.008
6 Initial 14.24 820.18 1630.55 0.8280 0.8002 0.733198 102.5
Final 8.29 924 1495 0.8442 0.8750 0.362168 8 0.041
7 Initial 10.5 1100 1693 0.76 0.77 2.483640 586
Final 8.3 921.5 1492.5 0.8446 0.8758 0.362136 28 0.032
8 Initial 6.1240 702.35 1522.54 0.8365 0.8216 0.531029 46.68
Final 8.43 911.5 1491.5 0.8480 0.8803 0.362053 9 0.009
9 Initial 12.5 900 1590 0.86 0.81 0.471676 30.29
Final 8.61 914.5 1492.5 0.8460 0.8778 0.362037 5 0.005
10 Initial 12.5 815 1392 0.75 0.75 0.797919 120
Final 8.52 914 1492 0.8470 0.8782 0.362033 11 0.003
Table 4
Variables for optimum solution from Ref. [5].
Variable PR T
3
(K) T
4
(K) g
AC
g
GT
Value 8.5234 914.28 1492.63 0.8468 0.8786
Table 5
Sum of the investment cost ow rate (
_
Z
T
), fuel cost ow rate (C

F
), total cost ow rate
(C

T
) and exergetic efciency (g
II
) corresponding to optimum solution from Ref [5].
_
Z
T
($/s)
C

F
($/s) C

T
($/s) g
II
(%)
0.036555 0.325465 0.362021 50.66
Fig. 5. Exergetic efciency (g
II
) with respect to total cost ow rate (C

T
) (numbers represent iteration).
2208 S.M. Seyyedi et al. / Energy Conversion and Management 51 (2010) 22022211
cases 4 and 9). Table 4 shows vector of decision variables for the
optimum conditions from Ref. [5]. Table 5 shows sum of the invest-
ment cost ow rate (
_
Z
T
), fuel cost ow rate (C

F
), total cost ow rate
(C

T
) and exergetic efciency (g
II
) corresponding to the optimum
conditions from Ref. [5] (see Appendix A).
6. Conclusions
A new approach has been developed for the optimization of
thermal power plant based on the exergoeconomic analysis and
structural optimization. In the real complex large thermal systems
that conventional mathematical optimization methods are not
efcient the proposed iterative methodology is efcient for
optimization. This methodology uses exergoeconomic analysis,
sensitivity analysis, and structural optimization method. The
advantages of this new iterative methodology are: (1) it can be ap-
plied to the real complex large thermal systems; (2) the procedure
of optimization is performed without user interface, i.e. it dose not
need to the decision of designer in each iteration; and (3) since it
uses from a numerical sensitivity analysis, its convergency is rapid.
The benets of the present work with respect to Refs. [2,23] can be
summarized as: (1) for each new iteration, new design variables
are calculated by structural optimization sub-routine (perfectly de-
scribed in Appendix B), while in Ref. [2] new design variables are
selected (and not calculated) by quality analysis of the component
with current design variable; (2) in large complex thermal systems,
(which have many design variables), it is difcult (or it is impossi-
ble) to determine which decision variable remains constant, in-
creases or decreases for the next iteration. Even if we overcome
on this problem, it is too difcult to determine the increased or de-
creased value of each decision variable; (3) since the present work
does not need any user-supplied data, this optimization can be
used widely while Ref. [2] needs to specialists of exergoeconomic
analysis for optimization; (4) a numerical sensitivity analysis is
used in the present work to achieve rapid convergency; (5) Ref.
[23] uses polyhedron method (a conventional mathematical opti-
mization method) for optimization in each iteration, while present
method uses structural optimization method which is more consis-
tent with thermal systems, because of the irreversibility rate and
investment cost rate terms in Eqs. (3); and (6) although numerical
sensitivity analysis in the present work and Ref. [23] is the same,
however, the aim of the making use of it is not the same. Compar-
ison of the present results with results of original CGAM problem
test shows suitable performance and good accuracy of the pro-
posed methodology.
Appendix A. Purchase costs of the component
When evaluating the costs of a plant, it is necessary to consider
the annual cost of fuel and the annual cost associated with owning
and operating each plant component. The expressions for obtain-
ing the purchase costs of the component (Z) are presented in Tables
A1 and A2. Based on the costs, the general equation for the cost
rate (
_
Z
i
in $/s) associated with capital investment and the mainte-
nance cost for the ith component is:
_
Z
i
Z
i
CRFu=N 3600 A:1
Here Z
i
is the purchase costs of the ith component ($), CRF is the
annual capital recovery factor (CRF = 18.2%), N represents the num-
ber of the hours of plant operation per year (N = 8000 h), and u is
the maintenance factor (u = 1.06).
In the CGAM problem the objective function is total cost ow
rate (C

T
) that is sum of the investment cost ow rate (
_
Z
T
) and fuel
cost ow rate (C

F
), i.e.
_
C
T

_
C
F

_
Z
T
A:2
where
_
C
F
_ m
f
c
f
LHV A:2:a
and
_
Z
T

m
k1
_
Z
k
A:2:b
where m is the number of components.
Appendix B. Develop structural optimization method for using
in the present new approach
B.1. Derive governing equation for structural optimization method
The purpose of this optimization is to determine for a selected
component (system element) the capital cost corresponding to
the minimum annual operating cost of the plant for a given plant
output and thus, by implication, to the minimum unit cost of the
product.
Assume that there is a plant parameter x
i
affecting the perfor-
mance of the kth element of the system and thus, in most case, also
indirectly affecting the performance of the system. Any variation in
x
i
will also, in general, cause changes in the irreversibility rates of
the other elements of the system, and necessitate changes in the
capital costs of the different elements. The exergy balance for the
system as a whole can be written:
_
I
T
x
i

_
E
IN
x
i

_
E
OUT
B:1
As shown, term E

OUT
which represents the joint exergy of the
plant products is taken to be independent of x
i
. The irreversibility
rate
_
I
T
(x
i
) may be looked upon as the consumption of exergy in
Table A1
Equations for calculating the purchase cost (Z) for the components.
Compressor
Z
AC

C11
_ ma
C12g
AC

P2
P1

ln
P2
P1

Combustion
chamber
Z
CC

C21
_ ma
C22
P
4
P
3

1 expC
23
T
4
C
24

Turbine
Z
GT

C31
_ mg
C32g
GT

ln
P4
P5

1 expC
33
T
4
C
34

Air preheater
Z
APH
C
41
_ mg h5h6
UDTLM

0:6
Heat-recovery
steam
generator
Z
HRSG
C
51
_
QPH
DTLM
PH

0:8

_
QEV
DTLM
EV

0:8

C
52
_ mst C
53
_ m
1:2
g
_ ma; _ mg ; _ mst are the mass ow rates of air, gas and steam respectively; h
5
and h
6
are
the specic enthalpies of streams 5 and 6; DTLM is the log mean temperature
difference;
_
Q
PH
and
_
Q
EV
represents the rate of heat transfer in the preheater
(economizer) and evaporator, respectively.
Table A2
Constants used in the equations of Table A1 for the purchase cost of the components
(Table A1).
Compressor C
11
39:5 $=kg=s C
12
0:9
Combustion chamber C
21
25:6 $=kg=s C
22
0:995
C
23
0:018 K
1
C
24
26:4
Turbine C
31
266:3 $=kg=s C
32
0:92
C
33
0:036 K
1
C
34
54:4
Air preheater C
41
2290 $=m
1:2
U 0:018 kW=m
2
K
Heat-recovery steam
generator
C
51
3650 $=kW=K
0:8
C
52
11820 $=kg=s
C
53
658 $=kg=s
1:2
S.M. Seyyedi et al. / Energy Conversion and Management 51 (2010) 22022211 2209
the system, necessary to generate the product exergy E

OUT
. An in-
crease in exergy consumption will necessitate corresponding to
additional exergy input, DE

IN
(x
i
).
The nature of the technique requires that the exergy input to the
plant should have a single xed unit cost. This condition can be sat-
ised by a single form of exergy input of invariable quality, e.g. fuel
or electric energy. Alternatively, the input could be made up of more
than one form of exergy of invariable quality in xed proportions.
The objective function is the total cost ow rate of operation for
the installation that is obtained from:
_
C
T
x
i
c
IN
_
E
IN
x
i

m
l1
_
Z
l
x
i
B:2
where C

T
is the total cost ow rate of fuel and equipment ($/s) and
_
Z
l
in ($/s) is the cost ow rate associated with capital investment
and the maintenance cost for the ith component.
Subject to the usual mathematical conditions being fullled, the
objective will be differentiated with respect to x
i
. From (B.1):
@
_
E
IN
@x
i

@
_
I
T
@x
i
B:3
So:
@
_
C
T
@x
i
c
IN
@
_
I
T
@x
i

m
l1
@
_
Z
l
@x
i
B:4
The second term on the RHS of (B.4) may be rearranged conve-
niently as:

m
l1
@
_
Z
l
@x
i

m
l
0
1
@
_
Z
l
0
@x
i

@
_
Z
k
@x
i
B:5
where l
0
k , i.e. subscript l
0
marks any of the element of the system
except that one which is subject to the optimization. Also, it will be
convenient to make the rearrangement:

m
l
0
1
@
_
Z
l
0
@x
i

@
_
I
k
@x
i

m
l
0
1
@
_
Z
l
0
@
_
I
k


@
_
I
k
@x
i
f
k;i
B:6
where
f
k;i

m
l
0
1
@
_
Z
l
0
@
_
I
k

x
i
var; l
0
k
B:7
f
k,i
is the capital cost coefcient.
The coefcient of structural bonds (CSB) is dened by:
r
k;i

@
_
I
T
@x
i

@
_
I
k
@x
i
B:8:a
Alternatively
r
k;i

@
_
I
T
@
_
I
k

x
i
var
B:8:b
From (B.8-a)
@
_
I
T
@x
i
r
k;i
@
_
I
k
@x
i

B:9
Using (B.5)(B.9), Eq. (B.4) can be modied as:
@
_
C
T
@x
i
c
I
k;i
@
_
I
T
@x
i

@
_
Z
k
@x
i
B:10
where
c
I
k;i
c
IN
r
k;i
f
k;i
B:11
For optimization, set equation (B.10) equal zero. Thus:
@
_
I
k
@x
i

OPT

1
c
I
k;i
@
_
Z
k
@x
i

B:12
B.2. Numerical solution of Eq. (B.12)
In this section, it has been described how Eq. (B.12) can be ap-
plied for structural optimization. Consider kth component and ith
decision variable. x
i
is ith decision variable that sets between x
iL
< -
x
i
< x
iU
with a step size Dx
i
Eq. (B.12) is numerically calculated using Newtons nite differ-
ence formula as:
_
I
k
x
i1

_
I
k
x
i

Dx
i


1
c
I
k;i
_
Z
k
x
i1

_
Z
k
x
i

Dx
i

B:13
where Dx
i
x
i1
x
i
. Consider:
f
1

_
I
k
x
i1

_
I
k
x
i

Dx
i

B:14
and
f
2

1
c
I
k;i
_
Z
k
x
i1

_
Z
k
x
i

Dx
i

B:15
Thus, when Eq. (B.13) is numerically evaluated, it can be written as:
f jf
1
f
2
j B:16
If f 0, then x
i
is obtained. Note that c
I
k;i
is numerically calcu-
lated for each x
i
. Fig. B1 shows functions f
1
and f
2
with respect to
x
i
(see Ref. [27]).
References
[1] Silveira JL, Tuna CE. Thermoeconomic analysis method for optimization of
combined heat and power systems. Part I Prog Energy Combust Sci
2003;29:47985.
[2] Tsatsaronis G, Moran M. Exergy-aided cost minimization. Energy Convers
Manage 1997;38(15):153542.
[3] Sahoo PK. Exergoeconomic analysis and optimization of a cogeneration system
using evolutionary programming. Appl Therm Eng 2008;28:15808.
[4] Bejan A, Tsatsaronis G, Moran M. Thermal design and optimization. New
York: Wiley; 1996.
f
2
f
1
x
i
x
Fig. B1. Functions f
1
and f
2
with respect to x.
2210 S.M. Seyyedi et al. / Energy Conversion and Management 51 (2010) 22022211
[5] Valero A, Lozano MA, Serra L, Tsatsaronis G, Pisa J, Frangopoulos CA, et al.
CGAM problem: denition and conventional solution. Energy 1994;19(3):
27986.
[6] Tsatsaronis G, Pisa J. Exergoeconomic evaluation and optimization of energy
systems: application to the CGAM problem. Energy 1994;19(3):287321.
[7] Frangopoulos CA. Application of thermoeconomic optimization methods to the
CGAM problem. Energy 1994;19(3):32342.
[8] Von Spakovsky MR. Application of engineering functional analysis and
optimization of the CGAM problem. Energy 1994;19(3):34364.
[9] Valero A, Serra L, Lozano MA, Torres C. Application of the exergetic cost theory
to the CGAM problem. Energy 1994;19(3):36581.
[10] El-Sayed YM, Gaggioli RA. A critical review of second law costing methods I:
background and algebraic procedures. ASME J Energy Resour Technol
1989;111:17.
[11] Gaggioli RA, El-Sayed YM. A critical review of second law costing methods II:
calculus procedures. ASME J Energy Resour Technol 1989;111:815.
[12] Lozano MA, Valero A. Theory of the exergetic cost. Energy Int J
1993;18(9):93940.
[13] Lazzaretto A, Tsatsaronis G. On the calculation of efciencies and costs in
thermal systems. In: Aceves SM et al. editors. Proceedings of the ASME
advanced energy systems division 1999, AES-vol. 39. New York: ASME; 1999.
p. 42130.
[14] Kim SM, Oh SD, Kwon YH, Kwak HY. Exergoeconomic analysis of thermal
systems. Energy 1998;23(5):393406.
[15] Kwon YH, Kwak HY, Oh SD. Exergoeconomic analysis of gas turbine
cogeneration systems. Energy Int J 2001;1(1):3140.
[16] Frangopoulos CA. Thermoeconomic functional analysis: a method for optimal
design or improvement of complex thermal systems. Ph.D. Thesis, Georgia
Institute of Technology, Georgia; 1983.
[17] Frangopoulos CA. Thermoeconomic functional analysis and optimization.
Energy 1987;12(7):56371.
[18] von Spakovsky MR. A practical generalized analysis approach to the optimal
thermoeconomic design and improvement of real-world thermal systems.
Ph.D. Thesis, Georgia Institute of Technology, Georgia; 1986.
[19] Erlach B, Serra L, Valero A. Structural theory as standard for thermoeconomic.
Energy Convers Manage 1999;40:162749.
[20] Hua B, Chen QL, Wang P. A new exergoeconomic approach for analysis and
optimization of energy systems. Energy 1997;22(11):10718.
[21] El-Sayed YM. A second-law-based optimization: part 1 methodology. ASME J
Eng Gas Turbines Power 1996;118:6937.
[22] Benelmir R, Feidt M. A comparative synthesis of exergo-economic
optimization: the IEEB method. In: Duncan AB et al. editors. Proceedings of
the ASME advanced energy systems division, AES-vol. 36. New York: ASME;
196. p. 44555.
[23] Vieira Leonardo S, Donatelli Joo L, Cruz Manuel E. Integration of an iterative
methodology for exergoeconomic improvement of thermal systems with a
process simulator. Energy Convers Manage 2004;45:2495523.
[24] Zhang C, Wang Y, Zheng C, Lou X. Exergy cost analysis of a coal red power
plant on structural theory of thermoeconomics. Energy Convers Manage
2006;47:81743.
[25] Leonardo S. Vieira, Joo L. Donatelli, Manuel E. Cruz. Exergoeconomic
improvement: an alternative to conventional mathematical optimization of
complex thermal systems. In: 6th World congress on structural
and multidisciplinary optimization. Rio de Janeiro, Brazil; 30 May03 June
2005.
[26] Leonardo S. Vieira, Joo L. Donatelli, Manuel E. Cruz. Integration of a
mathematical exergoeconomic optimization procedure with a process
simulator: application to the CGAM system. Engenharia Termica (Thermal
Engineering) 2005;4(2):16372.
[27] Kotas TJ. The exergy method of thermal plant analysis. Malabar (FL): Krieger
Pub.; 1995.
S.M. Seyyedi et al. / Energy Conversion and Management 51 (2010) 22022211 2211