Sunteți pe pagina 1din 14

Efciency enhancement of a gas turbine cycle using an optimized tubular

recuperative heat exchanger


Hoseyn Sayyaadi
*
, Reza Mehrabipour
Faculty of Mechanical Engineering-Energy Division, K.N. Toosi University of Technology, P.O. Box: 19395-1999, No. 15-19, Pardis Str., Mollasadra Ave., Vanak Sq.,
Tehran 1999 143344, Iran
a r t i c l e i n f o
Article history:
Received 28 February 2011
Received in revised form
19 November 2011
Accepted 22 November 2011
Available online 30 December 2011
Keywords:
Efciency enhancement
Payback time minimization
Multi-objective optimization
Bellman-Zadeh decision-making
LINMAP decision-making
TOPSIS decision-making
a b s t r a c t
A simple gas turbine cycle namely as the Kraftwerk Union AG unit including a Siemens gas turbine model
V93.1 with 60 MW nominal power and 26.0% thermal efciency utilized in the Fars power plant located
is considered for the efciency enhancement. A typical tubular vertical recuperative heat exchanger is
designed in order to integrate into the cycle as an air pre-heater for thermal efciency improvement.
Thermal and geometric specications of the recuperative heat exchanger are obtained in a multi-
objective optimization process. The exergetic efciency of the gas cycle is maximized while the
payback time for the capital investment of the recuperator is minimized. Combination of these objectives
and decision variables with suitable engineering and physical constraints makes a set of the MINLP
optimization problem. Optimization programming is performed using the NSGA-II algorithm and Pareto
optimal frontiers are obtained in three cases including the minimum, average and maximum ambient air
temperatures. In each case, the nal optimal solution has been selected using three decision-making
approaches including the fuzzy Bellman-Zadeh, LINMAP and TOPSIS methods. It has been shown that
the TOPSIS and LINMAP decision-makers when applied on the Pareto frontier which is obtained at
average ambient air temperature yields best results in comparison to other cases.
2011 Elsevier Ltd. All rights reserved.
1. Introduction
In this paper a simple gas turbine cycle namely as the Kraftwerk
Union AG unit utilized in the Fars gas power plant located in the
Shiraz city of Iran is considered for the efciency enhancement.
This unit is a Siemens gas turbine model V93.1 with 60 MW
nominal power and 26.0% thermal efciency at ISOcondition. There
are two categories of methods for efciency enhancement of gas
cycles. In the rst category the efciency of gas cycles is enhanced
using compressor inlet air cooling [1e3]. In this method, the inlet
air of the air compressor is cooled using evaporative coolers,
absorption chillers, electric chillers and similar apparatuses in
order to increase volumetric efciency of the compressor and
reducing compression work which leads to an increasing in the
efciency of the cycle and power generation. Second category of
methods dealing with increasing the combustion efciency and
therefore the efciency of the cycle is heat and gas recirculation
[4e8]. In the mass recirculation system, a portion of ue gas after
leaving the turbine is recirculated and mixed with the compressed
air coming from the air compressor or directly entered into the
combustion chamber. FGR (Flue gas recirculation) which some-
times called as EGR (exhaust gas recirculation) leads to both heat
recirculation by preheating of the inlet air of the combustion
chamber and dilution of the air-fuel mixture leading to less NO
x
formation. The FGR system systems offer the substantial reduction
of the thermal NO
x
formation (There is three mechanism of NO
x
formation including thermal nitrogen oxidation, prompt NO
x
and
fuel NO
x
[9]) due to reduction of N
2
and O
2
contents which are
substituted with CO
2
and H
2
O, shorter residence time of reactants
due to preheating and dilution, and reduction of local peak
temperatures due to better intermixing [4]. In the heat recircula-
tion systems the heat of combustion is recirculated to the inlet air of
the combustion chamber through the use of recuperative heat
exchanger or a regenerative one [4]. In heat recirculation the
thermal energy is transferred from combustion products into cold
substrates without mass transfer and thus without any dilution of
reactant [4]. Consequently, the total reactant enthalpy is increased
enabling sustained combustions. This leads to a self-sustained or
auto-thermal combustion that are sometimes referred to as
superadiabatic or excess enthalpy combustion [4]. The ideal of heat
recirculation is typical in the combustion science [4,5] and several
* Corresponding author. Tel.: 98 21 8867 4212; fax: 98 21 8867 4748.
E-mail addresses: sayyaadi@kntu.ac.ir, hoseynsayyaadi@gmail.com
(H. Sayyaadi).
Contents lists available at SciVerse ScienceDirect
Energy
j ournal homepage: www. el sevi er. com/ l ocat e/ energy
0360-5442/$ e see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.energy.2011.11.048
Energy 38 (2012) 362e375
combustion technologies such as mild combustion, recirculating
combustion have been developed based on this idea [6]. Moreover,
such approach can be benecial for a combustion process because
main exergy losses in the combustion arise from the heat transfer
from hot products to cold reactants while this temperature differ-
ence is benecially minimized in the gas turbine system involving
the air preheating [7]. Another technologies for heat recirculation is
the using a regenerative heat exchanger, however these equip-
ments are not usual in gas cycles but widely used as the air pre-
heatres of steam boilers namely as Ljungstrm. There are several
works involving heat recirculation approach for efciency
enhancement of gas turbines [5e8,10,11]. Ruixian et al. analyzed
the recuperative gas turbine cycle with a recuperator located
between HP and LP turbines namely as ARC (alternative recupera-
tion cycles) and compared it with the simple gas cycle and the CRC
(conventional recuperative gas cycle) [10]. They showed that the
efciency of this cycle might be higher than the CRC in the case of
same temperature ratio [10]. Further they indicated that the
maximum optimum efciency of practical ARCs is always lower
than that of CRCs and the optimum pressure ratio for efciency of
ARC is always higher than that of CRC.
Kim and Hwang analyzed part load performance of the recu-
perated gas cycle and specied which part load control strategy is
suitable for various congurations of the cycle [11].
In the current work, we consider integration of the air pre-heater
into the gas cycle for efciency enhancement of the proposed
Siemens gas turbine model V93.1. In this regard a typical tubular
vertical heat exchanger is designed as an air pre-heater of the
proposed gas cycle. Further we employ optimization approach for
integrating of the new component (air pre-heater) into the existing
Nomenclature
A
o
Heat transfer area (calculated based on tubes outside
diameter)(m
2
)
BL Booked life (years)
C Cost (US $)
CI Capital investment (US $)
CC
L
Levelized carrying charge (US $)
c Unit Cost (US $ per unit of the proposed parameter)
d
i
Distance of point ith from the ideal point
d
i
Distance of point ith from the non-ideal point
D Diameter (mm)
_
E The rate of exergy (kW)
e Specic exergy (kJ kg
1
)
e Specic molar exergy (kJ/kmol
1
)
F An objective function
h Heat transfer coefcient (W m
2
K
1
)
h Molar enthalpy (kJ kmol
1
)
i
eff
Interest rate (cost of money)
j jth year of operation
LHV Molar Lower Heat Value of fuel (kJ kmol
1
)
L
tp
Tube pitch in tube bundle (mm)
N
t
Number of tubes
N
b
Number of bafes
M Molecular weight (kg kmol
1
)
max Maximum operator in the fuzzy logics
min Minimum operator in the fuzzy logics
MOEA Multi-objective evolutionary algorithm
_ m Flow rate (kg s
1
)
_ n Molar ow rate (kmol.s
1
)
P Pressure (kPa)
payback Payback time for the capital investment
_
Q Heat transfer rate (kW)
r
FC
Annual escalation rate for the fuel cost
T Temperature (

C or K)
TRR
j
Total revenue requirement for jth year of the system
operation (US $)
s Molar specic entropy (kJ kmol
1
K
1
)
U
o
Overall heat transfer coefcient (calculated based on
outside tube surface) (W m
2
K
1
)
_
W Power (kW)
W Fabrication weight (kg)
DP Pressure difference or pressure drop (kPa)
Greek letters
r Density (kg m
3
)
h Efciency
n Specic volume (m
3
kg
1
)
m Membership function
Exergetic efciency
l fuel-air ratio (mass base)
l Molar fuel-air ratio
h
sc
Compressor isentropic efciency
h
sg
Gas turbine isentropic efciency
Subscripts
0 Index for ambient condition of the
atmospheric air, Index for the rst year
of the system operation
1,2,.,5,6 States 1,2,.,5,6 on regenerative gas cycle
a Air
ac Air compressor
bray Brayton gas cycle (simple gas cycle with
no air pre-heater)
cc Combustion chamber
duct Ducting for air and u gas transfer
to/from the recuperator
fab Fabrication
f Fuel
hx Heat exchanger
g Gas (ue gas)
gt Gas turbine
i ith element
it Tube inside
itl Inner tube limit in tube bundle
j jth element; jth year of the
system operation
L Levelized value
lm Log mean temperature difference (LMTD)
net Net
ot Tube outside
otl Outer tube limit in tube bundle
rec recuperator
reg Regenerative gas cycle
s Isentropic, shell side
t Tube side
stack Stack
Superscripts
max Maximum value for an objective
function
min Minimum value for an objective
function
n Non-dimensional
H. Sayyaadi, R. Mehrabipour / Energy 38 (2012) 362e375 363
gas cycle. Optimization is usually employed when a system is in
design stage; however we will show that in improvement stage of
existing systems optimization could be useful. Thermal and
geometric specications of the recuperative heat exchanger are
obtained in a multi-objective optimization process with the exer-
getic efciency and payback time of the recuperator investment as
objectives of optimization. Indeed multi-objective approach in
optimization of energy systems have been implemented by
researchers [12e18] to deal with energetic, economic and environ-
mental criteria, simultaneously. Energetic, economic and environ-
mental modeling of energy systems usually leads to a set of MINLP
(mixed integer non-linear) optimization problem. Therefore, in
order to avoid obtaining a local optimum instead of the global
optimum, meta-heuristic optimization approaches (e.g. genetic
algorithmor simulated annealing) have been utilized by researchers
nowadays [13,15,18e20]. Chang and Hwang [13] developed a MINLP
multi-objective model for an energy system to reduce the cost and
environmental impact. Roosen et al. [17] implemented multi-
objective optimization to balance between the capital investment
and operating cost of a combined power cycle. Multi-objective
optimization of a benchmark cogeneration system namely as
CGAM problemwas conducted by one of authors [13]. In that work,
three objective functions including the exergetic, the total levelized
cost rate of the system product and the cost rate of the environ-
mental impact were considered simultaneously.
In a most relevant work Sayyaadi et al. designed a particular
non-TEMA type recuperative heat exchanger utilized as a regener-
ator of a regenerative gas turbine cycle, for a gas cycle [21]. They
considered a tubular heat exchanger with a vertical annular tube
bundle and disk and doughnut types bafes. The exergetic ef-
ciency of the entire gas cycle together with the purchased equip-
ment cost of its recuperative heat exchangers was considered as
objectives of optimization. It was assumed that the recuperator is
designed for an existing gas turbine cycle to be retrotted. Three
scenarios for optimization of the system including the minimiza-
tion of the recuperator cost, maximizing the cycle exergetic ef-
ciency and simultaneous optimization of both objectives were
performed. An example of decision-making was presented to select
a nal optimal solution from the Pareto frontier. Finally the results
which were obtained by three optimization scenarios were
compared together and also compared with the base case system.
In this work, the cost of recuperator which was the second
objective functionof the ref. [21] is substitutedwiththe payback time
of the recuperator investment. Therefore, the payback time of the
recuperator is minimized instead of the recuperator cost. This new
objective function is comprised from the capital investment of the
recuperator and the fuel cost of the gas cycle. Therefore, it means that
in the current research the three objectives including the exergetic
efciency, the cost of the recuperator and the saving of the fuel cost
are optimized. The last two objectives are integrated in the payback
time objective in fact. Therefore, minimizing the payback time of the
recuperator investment leads to either minimizing the recuperator
cost or maximizing the fuel cost saving through using of the recu-
perative gas cycle. Hence, in this work one more objective beside
those objectives which were considered in ref. [21] is optimized.
Further the economic model for evaluation of the carrying charge
and fuel cost is modied based on the value of interest rate and fuel
escalation factor; hence levelized values for capital investment and
fuel cost are employed in evaluation of the payback time. Combina-
tion of aforementioned objectives and decision variables including
tubes length, tubes outside/inside diameters, tube pitch in the tube
bundle, outer and inner tube limits of the tube bundle and the total
number of disc and doughnut bafes plus the air outlet temperature
from the recuperator with suitable engineering and physical
constraints makes a set of the MINLP optimization problem.
Optimizationprogramming inMATLABis performedusingone of the
most powerful and robust multi-objective optimization algorithms
namely as the NSGA-II. Since operation of the gas cycle is highly
dependent to the ambient air temperature, the multi-objective
optimization is performed in three cases of ambient condition
including the minimum, average and maximum annual condition at
the site of gas cycle (Shiraz city in Iran). Further, in additional
improvement on previous work [21], three decision-making
approaches including the fuzzy Bellman-Zadeh [22], LINMAP
[23,24] and TOPSIS [23,24] are utilized for selection of nal optimal
solutions from the Pareto frontiers which obtained at the minimum,
average and maximumenvironmental temperature. Therefore, three
optimal solutions are selected using aforementioned decision-
makers at three ambient cases (minimum, average and maximum
annual air temperature at the site). Since the performance of the gas
cycle is signicantly affected by the ambient air condition, the
question is which ambient condition should be considered as the
best design condition. In this paper we try to answer this question
and we will present a systematic method to specify which ambient
temperature should be taken account for designing of the regener-
ative gas cycle.
2. Problem denition
As is previously mentioned in this paper, the proposed gas
turbine cycle is a simple Brayton gas cycle. This gas cycle is
a Siemens unit model V93.1 namely as the Kraftwerk Union AG unit
with 60 MW nominal power and 26.0% thermal efciency at ISO
condition (15

C ambient air temperature at sea level). One Siemens


V93.1 gas turbine has been installed in the Fars gas power plant
located in the Shiraz city, Iran and has been in operation since 1980.
As the unit is an old unit with a relatively low efciency in
comparison to current technologies of gas turbines, the objective of
this project is enhancement of the proposed V93.1 unit as much as
possible in order to operate it with a more reasonable efciency.
Table 1 show general specications of the V93.1 Brayton gas cycle.
In this paper, the thermal efciency of the proposed gas turbine
will be enhanced by integration of a recuperative heat exchanger as
an air pre-heater. Fig. 1 shows a schematic arrangement of the
proposed regenerative gas turbine cycle with a recuperative heat
exchanger as an air pre-heater. Combustion chamber inlet air is
pre-heated using the ue gas exhausts from the gas turbine.
The recuperative heat exchanger that will be integrated into the
gas cycle as an air-preheater comprises of a vertical tubular shell
and tube heat exchanger that directly is connected to a conical stack
at the top of the heat transfer area. The compressed air at the outlet
of the air compressor enters to the shell side of the recuperator
from the top and is pre-heated by the ue gas that enters in tubes
from the bottom of heat exchanger. The pre-heated air exits from
the bottom of shell side and directed to the combustion chamber.
Flue gas exits from tubes at the top and directed into the stack
section (conical section at the top of heat transfer area). The
proposed heat exchanger has an annular tube bundle with disk and
doughnut bafes. More detail on specications of the proposed
recuperative heat exchanger can be found in [21].
The aim is nding the geometrical specications of the recu-
perative heat exchanger including tube length, outside/inside
diameters of tubes, tube pitch in the tube bundle, outer and inner
tube limits of tube bundle and the number of bafes plus the pre-
heated air outlet temperature from the recuperator so that the
exergetic efciency of the cycle is maximized and the payback time
for capital investment of the recuperator is minimized,
simultaneously.
H. Sayyaadi, R. Mehrabipour / Energy 38 (2012) 362e375 364
3. System modeling
3.1. Thermodynamic modeling
Thermodynamic model of the entire regenerative gas turbine
cycle is built based on the following basic assumptions
1. All processes are steady state.
2. The principle of ideal-gas mixtures is applied for the air and
combustion products.
3. The fuel is the natural gas and it is assumed to be 100% methane
and ideal gas.
4. Heat loss from the combustion chamber is considered to be 2%
of the fuel lower heating value. All other components are
assumed adiabatic.
5. Constant pressure loss ratios are considered in the system
components except in the recuperator (pressure losses in
recuperator are calculated based on hydraulic calculations).
6. Molar fractions for composition of the inlet air are assumed to
be 0.7594N
2
, 0.2038 O
2
, 0.0003 CO
2
and 0.0274H
2
O.
7. Isentropic efciencies of the air compressor and gas turbine are
assumed constant.
Therefore, thermodynamic equation of the cycles (simple and
recuperative gas cycles) are developed as follows,
3.1.1. Air compressor
In order to achieve the outlet isentropic temperature of the air
compressor we have:
T
2s
T
1
= r
k1
k
pc
(1)
Where r
pc
is the compression ratio of the compressor (=8.7). The
real outlet temperature of the compressor is:
T
2
=
_
T
2s
T
1
h
SC
_
T
1
(2)
Isentropic efciency of the compressor,h
SC
, is 0.84. Assuming an
adiabatic compressor, the consumed power of the air compressor
is,
_
W
ac
= _ m
a
(h
2
h
1
) (3)
Fig. 1. Schematic for a regenerative gas turbine cycle.
Table 1
Specications of the simple gas turbine.
Manufacturer Kraftwerk
union AG
Type Siemens V93.1
Number of turbine stages 4
Rotor speed (rpm) 3000
Air ow rate at ISO condition (kg s
-1
) 343.4
Flue Gas ow rate with gas oil at ISO condition (kg s
-1
) 348.6
Flue Gas ow rate with natural gas at ISO condition (kg s
-1
) 347.8
Turbine inlet temperature for the base load operation
at rated output (

C)
850
Turbine inlet temperature for the peak load operation
at rated output (

C)
870
Compressor type Single ow axial
type V 93.1
Number of compressor stages 16
Compressor air ow rate at the ISO condition (kg s
-1
) 343.4
Compression ratio of the compressor at the ISO condition 8.70
Compression ratio of the turbine at the ISO condition 8.35
Isentropic efciency of the air compressor 0.84
Isentropic efciency of the turbine 0.85
Combustion chamber type Vertical silo type
Number of combustors 4
Pressure loss in the combustion chamber 2% of the
inlet pressure
H. Sayyaadi, R. Mehrabipour / Energy 38 (2012) 362e375 365
3.1.2. Recuperator
Detail thermohydraulic formulas and thermodynamic correla-
tion for modeling of the recuperator have been presented in [21].
3.1.3. Combustion process
Chemical reaction equation of the reaction process can be
formulated as,
lCH
4

_
x
N2;a
N
2
x
O2;a
O
2
x
CO2;a
CO
2
x
HO2;a
H
2
O

/[1 l[

_
x
N2;g
N
2
x
O2;g
O
2
x
CO2;g
CO
2
x
HO2;g
H
2
O

(4)
Where l is the molar fuel to air ratio and
x
N2;g
=
x
N2;a
1l
; x
O2;g
=
x
O2;B
2l
1l
; x
CO2;g
=
x
CO2;B
l
1l
; x
HO2;g
=
x
HO2;B
2l
1l
(5)
In Eq. (5) subscripts a and g denote the property of species
(molar composition) in the atmospheric air and ue gas, respec-
tively. The energy balance for the combustion chamber is:
0 =
_
Q
CV

_
W
CV
_ n
f
h
f
_ n
a
h
a
_ n
p
h
p
(6)
Where index f presents the fuel, index a points to the air and
index p points to products of the combustion. In Eq. (6) we have
_
W
CV
= 0 and since we assumed the heat loss fromthe combustion
chamber to be 2% of LHV of the fuel,
_
Q
CV
= 0:02 _ n
f
LHV = 0:02 _ n
a
l LHV (7)
Therefore Eq. (6) is converted to the following form,
0 = 0:02lLHV
f
h
a
lh
f
(1 l)h
p
(8)
LHV
f
for the methane as a fuel is 3124 kJ kmol
1
. Further we
have:
h
a
=
_
x
N2;a
h
N2
x
O2;a
h
O2
x
CO2;a
h
CO2
x
HO2;a
h
H2O

at T3
(9a)
h
p
=
_
x
N2;g
h
N2
x
O2;g
h
O2
x
CO2;g
h
CO2
x
HO2;g
h
H2O

at T6
(9b)
Hence l is obtained from solution of Eq. (8), therefore, the mass
ow rate of the fuel is calculated as,
_ n
F
= l _ n
a
/_ m
F
= l
_
M
f
M
a
_
_ m
a
(10)
Where M
f
and M
a
are molecular weights of the fuel and air,
respectively.
3.1.4. Gas turbine
In similar way to the air compressor, the isentropic outlet
temperature of the gas turbine is determined as,
T
4
T
6s
= r
k1
k
pg
(11)
Where r
pg
is the expansion ratio of the gas turbine (=8.35) and k =
c
p
c
p
R
in which R =
R
M
t
. M
t
is the molecular weight of outlet gas.
Therefore,
T
5
= T
4
h
sg
(T
4
T
5s
) (12)
The isentropic efciency,h
sg
, of the gas turbine is 0.85 here.
Considering the turbine as a control volume and an adiabatic
turbine, from energy balance we have:
_
W
gt
= (1 l) _ n
a
(h
4
h
5
) = _ m
p
(h
4
h
5
) (13)
3.1.5. Exergetic efciency of the gas cycle
The exergetic efciency of the gas cycle is determined as follows,

tot
=
_
W
net
_ m
f
e
ch
f
=
_
W
gt

_
W
ac
_ m
f
e
ch
f
(14)
Where
_
W
net
is the net generated power and e
ch
f
is the chemical
exergy of the fuel assumed as 53155.8 kJ kg
1
for methane.
3.2. Thermohydraulic modeling of the recuperative heat exchanger
Thermohydraulic model is used here in order to calculate the
required heat transfer area for the recuperative heat exchanger in
the one hand and calculating of the hot and cold streams pressure
drops which affect performance of the gas cycle on the other hand.
An especial type of the shell and tube heat exchanger with a vertical
annular tube bundle and disk and doughnut shape bafes is used
here as the air pre-heater of the gas cycle. Complete thermohy-
draulic model for this type of heat exchanger was presented by
Sayyaadi et al. in [21].
3.3. Economic modeling
As is previously mentioned, the payback time for the capital
investment of the recuperative heat exchanger is considered as the
secondary objective of this work. Total capital investment of the
recuperator is comprised from the cost of heat transfer area plus
the cost of stack section and the piping cost. Therefore, the capital
investment of the gas cycle enhancement is,
CI = C
hx
C
stack
C
duct
(15)
The capital investment of a heat exchanger (the heat transfer
area) can be estimated using the following expression [25],
C
hx
= 8500 409A
0:85
o
(16a)
Where A
o
is the total outside area of the tube bundle which is
calculated using the thermohydraulic model of the recuperative
heat exchanger [21]. The cost of stack section is obtained based on
the fabricated weight of the stack and the manufacturing cost per
kg of the fabricated weight as follow,
C
stack
= c
fab
W
stack
(16b)
W
stack
is the fabricated weight of the stack section obtained in
mechanical design, c
fab
is the fabricated price of the stack per kg of
its weight taken as 3.0 $.kg
1
(in the Iran). In Eq. (18) the ducting
cost, C
duct
, based on the site investigation is considered as 38,500$
in this case (ducts are used in order to transfer the air and ue gases
to/from the recuperator).
A levelized value for the total annual cost of the capital
investment, CC
L
, can be computed by applying a discounting
factor (the cost of money or interest rate) and the capital-recovery
factor CRF:
CC
L
= CRF

BL
1
TRR
j
_
1 i
eff
_
j
(17a)
H. Sayyaadi, R. Mehrabipour / Energy 38 (2012) 362e375 366
CRF =
i
eff
_
1 i
eff
_
BL
_
1 i
eff
_
n
1
(17b)
Where TRR
j
is the total revenue requirement in the jth year of the
system operation [13] obtained as,
TRR
j
=
CI
BL
(17c)
In Eqs (17aec) i
eff
is the average annual effective discount rate
(cost of money), and BL denotes the system economic life (booked
life) expressed in years. The operating life of the system is assumed
to be 20 years and i
eff
in Iran is 17%.
Therefore, the payback time for the capital investment of the
recuperator is calculated based on the levelized capital investment,
CC
L
, (Eq. (17a)) and the annual saving on the fuel cost reduction as
follow,
payback =
CC
L
BL
c
fL
86400365
_
1
=
r
f
_

_
_ m
f
Simple cycle
_ m
f
Recuperative cycle
_
(18)
Where c
fL
is the levelized cost of each cubic meters of the natural
gas. If the series of payments for the annual fuel cost is uniformover
time except for a constant-escalation r
FC
then the levelized value for
the fuel cost per cubic meters, c
fL
, of the series can be calculated by
multiplying the unit cost of fuel at the beginning of the rst year of
the project c
f0
by the constant-escalation levelization factor CELF as
follows
c
fL
= c
f0
CELF = c
f0
k
FC
_
1 k
BL
FC
_
(1 k
FC
)
CRF (19a)
With
k
FC
=
1 r
FC
1 i
eff
and r
FC
= constant: (19b)
The terms r
FC
and CRF denote the annual escalation rate for the
fuel cost (assumed to be 5%) and the capital-recovery factor (Eq.
(17b)), respectively.
In Eq. (19a) c
f0
is 0.08 $ m
3
in Iran (the international price for
natural gas is approximately 0.30 $ m
3
about 3.75 times higher
than the local price in Iran).
3.4. Model verication
Thermodynamic and thermohydraulic modeling of the cycle is
performed using the MATLAB programming. For verication of the
model, the registered data at the site of the proposed gas turbine for
the exhaust temperature of the gas is compared with correspond-
ing data predicted by the thermodynamic model. The thermody-
namic model has been performed for site condition with 84.7 kPa
atmospheric pressure and the ambient registered temperature.
Fig. 2 illustrates this comparison for 27 days selected during an
operating year 2009, randomly.
This gure indicates that the maximum error is 2.2%. Further
operation of the simple gas cycle at ISO condition predicted by the
developed model is compared with the system catalogue (reported
by the manufacturer) as indicated in Table 2.
As is clear, Table 2 indicates that in this case the maximum error
is also 2.4%. Therefore, our thermodynamic model is able to predict,
the gas cycle behavior with a maximum 2.4% error which is
reasonable for our purpose.
4. Objective functions, decision variables and constraints
4.1. Denition of the objectives
As is already discussed, in this paper two objectives including
the exergetic efciency of the regenerative gas turbine and the
payback time for the capital investment of recuperator denoted by
Eqs. (14) and (18), respectively, are considered. The exergetic ef-
ciency (Eq. (14)) is maximized while the recuperator investment
payback time (Eq. (18)) is minimized. These objectives are consid-
ered simultaneously in a multi-objective optimization process.
4.2. Choice of decision variables
Following geometrical and thermal specications of the recu-
perative heat exchanger are considered as decision variables,
v L
t
: Tube length (m)
v D
to
: Tube outside diameter (m)
v D
ti
: Tube inside diameter (m)
v L
tp
: Tube pitch in the tube bundle (center to center distance of
tubes in m)
v D
otl
: Outer tube limit in the tube bundle(m)
v D
itl
: Inner tube limit in the tube bundle (m)
v N
b
: Total number of bafes (including disk and doughnut
bafes)
v T
3
: The outlet temperature of the pre-heated air from the
recuperator (K)
4.3. Constraints and limitations
Following limitations are considered for the regenerative gas
turbine cycle:
Fig. 2. Comparison of the exhaust gas temperatures predicted by the model and real
values.
Table 2
Comparison of the performance of the simple gas cycle at the ISO condition pre-
dicted by the thermodynamic model and catalogue reported values.
Specication Catalogue data Predicted by the code Error
Efciency 0.26 0.266 2.3%
Air ow rate (kg s
1
) 343.4 335.3 2.4%
Fuel ow rate (kg s
1
) 4.4 4.49 2.0%
H. Sayyaadi, R. Mehrabipour / Energy 38 (2012) 362e375 367
3 _ v
air
_ 6m:s
1
(20)
Where v
air
is the air velocity in the circular cross section area of
doughnut bafes and the annular area limited between the outer
tube limit of the tube bundle and shell inside area.
T
3
_ 1420 K (21)
T
6
_ 378:15 K (22)
6 _ L
t
_ 12 m (23)
L
tp
D
to
1:25; 1:33; 1:5 (24)
N
b
3; 5; 7; 9; 11; 13; 15 (25)
2 _ D
si
_ 4 m (26)
0:80 _
D
otl
D
si
_ 0:97 (27)
0:25 _
D
itl
D
si
_ 0:45 (28)
T
3
< T
4
< T
5
(29)
DP
s
_ 3:5kPa (30)
DP
t
_ 5kPa (31)
Since implementing of the air pre-heater in some case may
cause efciency reduction of the gas cycle even lower than the
simple gas turbine with no air pre-heater, in order to avoid such un
reasonable condition, the following constraint has been imposed
on the optimization process,

reg
_
bray
0:01 (32)
Where
reg
and
bray
are exergetic efciencies of the regenerative
and simple Brayton gas cycles, respectively.
5. Multi-objective optimization
Multi-objective optimization of objectives function expressed
by Eqs. (14) and (17) is performed using the multi-objective
evolutionary algorithm. A multi-objective optimization problem
requires the simultaneous satisfaction of a number of different and
often conicting objectives. It is required to mention that no
combination of decision variables can optimize all objectives,
simultaneously. Multi-objective optimization problems generally
show a possibly uncountable set of solutions, whose evaluated
vectors represent the best possible trade-offs in the objective
function space. Pareto optimality is the key concept to establish
a hierarchy among the solutions of a multi-objective optimization
problem, in order to determine whether a solution is really one of
the best possible trades-off [26]. Eq. (1) shows how a multi-
objective optimization problem can be formulated mathematically.
min F
j
(X) cj1; 2; 3; .:; k subject to XL (33)
Where we have k _ 2 objective functionsF
j
: R
n
/R
1
. The feasible
objective region Z is the image of the feasible region (i.e Z = F (X)
3R
k
). The elements of Z are called objective vectors. The objective
vectors are denoted by F(X) or by Z =[z
1
,z
2
,z
3
.,z
k
]
T
, where z
j
=F
j
(X)
cj {1,2,.,k} [27].
Classical search and optimization methods are not efcient in
following the Pareto approach for multi-objective optimizations.
The class of search algorithms that implement the Pareto approach
for multi-objective optimization in the most straightforward way
is the class of multi-objective evolutionary algorithms (MOEAs)
[12]. In this paper, one of most powerful MOEA namely as Non-
dominated sorting genetic algorithm, NSGA-II has been
employed to nd the Pareto optimal frontier for the proposed
recuperative gas cycle. This method was well described by
Sayyaadi et al. in [21].
6. Decision-making in the multi-objective optimization
In multi-objective optimization a process of decision-making
for selection of the nal optimal solution from available solutions
is required. There are several methods for decision-making process
in decision problem. These methods can be employed for selection
of a nal optimal solution from the Pareto frontier Since, dimen-
sion of various objectives in a multi-objective optimization
problem might be different (for example in our case the exergetic
objective has no dimension while the dimension of the payback
time is in years), therefore, before any decision, dimension and
scales of objectives space should be unied. In this regard, objec-
tives vectors should be non-dimensionalized before decision-
making. There are some methods of non-dimensionalization
utilized in decision making including linear non-
dimensionalization, Euclidian non-dimensionalization, and fuzzy
non-dimensionalization.
v Linear non-dimensionalization
Consider the matrix of objectives at various points of the Pareto
frontier is denoted by F
ij
where i is the index for each point on the
Pareto frontier and j is the index for each objective in the objec-
tives space. Therefore a non-dimensionalized objective, F
n
ij
, is
dened as,
F
n
ij
=
F
ij
max
_
F
ij
_ for maximizing objectives (34a)
F
n
ij
=
1=F
ij
max
_
1=F
ij
_ for minimizing objectives (34b)
v Euclidian non-dimensionalization
In this method, a non-dimensionalized objective, F
n
ij
, is dened
as,
F
n
ij
=
F
ij

m
i=1
_
F
ij
_
2 2
_ for minimizing and maximizing objectives
(35)
v Fuzzy non-dimensionalization
In this method, a non-dimensionalized objective, F
n
ij
, is dened
as,
H. Sayyaadi, R. Mehrabipour / Energy 38 (2012) 362e375 368
F
n
ij
=
F
ij
min
_
F
ij
_
max
_
F
ij
_
min
_
F
ij
_ for maximizing objectives (36a)
F
n
ij
=
max
_
F
ij
_
F
ij
max
_
F
ij
_
min
_
F
ij
_ for minimizing objectives (36b)
In this paper most famous and common type of decision-
making processes including the fuzzy Bellman-Zadeh, LINMAP
and TOPSIS method is used in parallel in order to specify the nal
optimal solution. The Bellman-Zadeh method utilizes the fuzzy
non-dimensionalization while the other methods (LINMAP and
TOPSIS) employ Euclidian non-dimensionalization. The following
sections are presented here in order to describe these decision-
making algorithms.
6.1. Bellman-Zadeh decision-making method
When using the Bellman-Zadeh approach, each F
j
(X) of Eq. (33)
is replaced by a fuzzy objective function or a fuzzy set
A
j
=
_
X; m
A
j
(X) XL; j = 1; 2; .:k (37)
Where m
Aj
(X) is a membership function of A
j
[28].
A nal decision is dened by the Bellman and Zadeh model as
the intersection of all fuzzy criteria and constraints and is repre-
sented by its membership function. A fuzzy solution D with setting
up the fuzzy sets (38) is turned out as a result of the intersection
D = X
k
j=1
A
j
with a membership function
m
D
(X) = X
k
j=1
m
A
j
(X) = min
j =1;.:;k
m
A
j
(x) XL (38)
Using Eq. (35), it is possible to obtain the solution proving the
maximum degree as follows,
maxm
D
(X) = max
XL
min
j =1;.:;k
m
A
j
(x) (39)
X
0
= argmax
XL
min
j =1;.:;k
m
A
j
(x) (40)
To obtain Eq. (39), it is necessary to build membership
functions m
Aj (X)
, j = 1,., k reecting a degree of achieving own
optima by the corresponding F
j
(X); XL, j = 1, . . . , k. This is satised
by the use of the membership functions [24]. The membership
function of objectives and constraints, linear or non-linear, can be
chosen depending on the context of problem. One of possible fuzzy
convolution schemes is presented below [25].
v Initial approximation for X-vector is chosen. Maximum
(minimum) values for each criterion F
j
(X) are established via
scalar maxi mization (minimization). Results are denoted as
ideal points X
0
j
; j = 1; .; m.
v The matrix table {T}, where the diagonal elements are ideal
points, is dened as follows:
T =
_

_
F
1
_
X
0
1
_
F
2
_
X
0
1
_
. F
n
_
X
0
1
_
F
1
_
X
0
2
_
F
2
_
X
0
2
_
. F
n
_
X
0
2
_
:
:
:
F
1
_
X
0
n
_
F
2
_
X
0
n
_
. F
n
_
X
0
n
_
_

_
(41)
v Maximum and minimum bounds for the criteria are dened:
F
min
i
= min
j
F
j
_
X
0
j
_
; i = 1; .:; n
F
max
i
= max
j
F
j
_
X
0
j
_
; i = 1; .:; n
(42a,b)
v The membership functions are assumed for all fuzzy goals as
follows. For minimized objective functions
m
Fi
(X) =
_

_
0 if F
i
(x)>F
max
i
;
F
max
i
F
i
F
max
i
F
min
i
if F
min
i
< F
i
_ F
max
i
;
1 if F
i
(x) _ F
min
i
(43a)
For maximized objective functions
m
Fi
(X) =
_

_
1 if F
i
(x)>F
max
i
;
F
i
F
min
i
F
max
i
F
min
i
if F
min
i
< F
i
_ F
max
i
;
0 if F
i
(x) _ F
min
i
(43b)
v Fuzzy constraints are formulated:
G
j
(X) _ G
max
j
d
j
; j = 1; 2; .:; k (44)
Where d
j
is a subjective parameter that denotes a distance of
admissible displacement for the bound G
max
j
of the jth constraint.
Corresponding membership functions are dened in following
manner:
m
Gi
(X) =
_

_
0 if G
i
(x)>G
max
i
1
G
j
(x) G
max
j
d
j
if G
max
i
<G
i
(X) _G
max
i
d
j
1 if G
i
(x) _G
max
i
(45)
v A nal decision is determined as the intersection of all fuzzy
criteria and constraints represented by its membership func-
tions. This problem is reduced to the standard non-linear
programming problems: to nd the such values of X and k
that maximize k subject to
l _ m
F
i
; i = 1; 2; .:; n
l _ m
G
j
; j = 1; 2; .:; k
(46)
The solution of the multi-criteria problemdiscloses the meaning
of the optimality operator and depends on the decision-makers
experience and problem understanding.
H. Sayyaadi, R. Mehrabipour / Energy 38 (2012) 362e375 369
6.2. LINMAP decision-making method
An ideal point on the Pareto frontier is the point in which each
objective is optimized regardless to satisfaction of other objectives.
It is clear that in the multi-objective optimization it is impossible to
have each objective in its optimal condition obtained a single-
objective optimization. Therefore, the ideal point is not located
on the Pareto frontier. In the LINMAP method, after Euclidian non-
dimensionalization of all objectives, the spacial distance of each
solution on the Pareto frontier from the ideal point denoted by d
i
is determined as follow,
d
i
=

n
j=1
_
F
ij
F
Ideal
j
_
2
2
_
(47)
Where n denotes the number of objective while i stand for each
solution on the Pareto frontier (i = 1,2,..,m). In Eq. (47),F
Ideal
j
is the
ideal value for jth objective obtained in a single-objective optimi-
zation. In LINMAP method, the solution with a minimum distance
from the ideal point is selected as a nal desired optimal solution,
hence, i index for a nal solution, i
nal
is,
i
final
himin(d
i
) i = 1; 2; .; m (48)
6.3. TOPSIS decision-making method
In this method beside the ideal point a non-ideal point is
dened. The non-ideal point is the ordinate in the objectives space
in which each objective has its worst value. Therefore, beside the
solution distance from ideal point, d
i
, the solution distance from
the non-ideal point denoted by d
i
is used as a criterion for selec-
tion of the nal solution. Hence,
d
i
=

n
j=1
_
F
ij
F
Nonideal
j
_
2
2
_
(49)
In continuing the TOPSIS method a Cl
i
parameter is dened as
follows,
Cl
i
=
d
i
d
i
d
i
(50)
In the TOPSIS method a solution with a maximum Cl
i
is selected
as a desired nal solution, therefore, if i
nal
is index for the nal
selected solution, we have,
i
final
himax(Cl
i
)i = 1; 2; .; m (51)
7. Results and discussion
The present simple gas cycle is modeled using the thermody-
namic model for the site minimum, average and maximum annual
temperatures. Table 3 indicates the simple gas cycle performance at
aforementioned site conditions.
Now a recuperative heat exchanger is integrated to the V93.1
gas cycle in order to convert it into a regenerative gas cycle that
schematically is shown in Fig. 1. In this regards, geometric and
thermal specications of the recuperative heat exchanger and the
air outlet temperature from the recuperator are specied in
a multi-objective optimization process with objective functions
expressed By Eqs. (14) and (18) and constraints specied by Eqs.
(21)e(32). Multi-objective optimization using NSGA-II algorithm
is performed in three site cases including annual minimum,
average and maximum ambient air temperatures and Pareto
frontiers are obtained in these three cases as illustrated in
Fig. 3(aec), respectively. In each case, nal optimal solutions
have been selected with fuzzy, LINMAP and TOPSIS decision-
makers.
It is clear from Fig. 3b that at the average annual site tempera-
ture, LINMAP and TOPSIS recommend the same nal optimal
solution. Tables 4e6 indicates specications of the regenerative
cycles that recommended by fuzzy, LINMAP and TOPSIS decision-
makers.
Nowthe question is that which decision-making method should
be considered for a nal selection. Table 7 and Fig. 3(aec) are
presented here to help us to decide between various decision-
making methods. In the Table 7, it is assumed that the system
optimized in each temperature (optimization base temperature) is
operated in other two ambient temperatures and the efciency
improvements in those conditions are calculated. From the last
column of Table 7, we can see that all decision-makers lead to
approximately the same values for the average exergetic efciency
(1.63%, 1.57% and 1.61% for fuzzy, LINMAP and TOPSIS decision-
makers, respectively).
Fig. 4 shows that, in almost same exergetic efciency improve-
ment of three decision-makers, LINMAP and TOPSIS decision-
makers lead to the lowest payback time for the recuperator
investment. Therefore, it seems that in this case, LINMAP and
TOPSIS provide a more desirable nal optimal solution.
It should be mentioning that in general there is no decision-
making method having superiority over other methods in all
cases. Indeed, various decision-making methods are applied to help
decision-makers who select the nal solution based on their
professional experience. In this case, we applied three decision-
making methods and we found that in our case LINMAP and TOP-
SIS decision-makers select a nal optimal solution that more suit
our engineering and economic criteria.
Another question is which ambient annual temperature should
be considered as a reference base temperature for optimization of
the system. In this regard, once more we assumed that the system
optimized at each ambient temperature is operated in other two-
ambient temperatures. For example we assumed that the system
optimized at the minimum temperature is utilized in the average
and maximum ambient temperatures. Now, the deviations of the
real system (obtained with 3.5

C base temperature) from the
optimal solution at 18.8

C and 40.6

C are assessed. This procedure
is more elucidated as follows,
d
18:8
+
C

_
p
3:5
+
C
n
p
18:8
+
C
n
_
2

3:5
+
C
n

18:8
+
C
n
_
2
_
(52a)
d
40:6
+
C

_
p
3:5
+
C
n
p
40:6
+
C
n
_
2

3:5
+
C
n

40:6
+
C
n
_
2
_
(52b)
Table 3
Performance of the simple Brayton V93.1 gas cycle at site conditions (Fars gas power
plant located in Shiraz-Iran).
Ambient
temp. (

C)
Ambient
pressure (kPa)
Net power
(kW)
Exergetic
efciency (%)
Fuel mass ow
rate (kg s
1
)
T
min
= 3.5 84.7 55683 26.36 3.9865
T
ave
= 18.8 84.7 48160 24.83 3.6610
T
max
= 40.6 84.7 40664 22.93 3.3472
H. Sayyaadi, R. Mehrabipour / Energy 38 (2012) 362e375 370
Fig. 3. Pareto optimal frontiers at (a) Minimum annual temperature; (b) Average annual temperature; (c) Maximum annual temperature.
H. Sayyaadi, R. Mehrabipour / Energy 38 (2012) 362e375 371
Where d
18:8
+
C

and d
40:6
+
C

are deviations of the optimal solution


obtained at 3.5

C from optimal solutions obtained at 18.8

C and
40.6

C, respectively. p
n
and
n
denote Euclidian non-
dimensionalized payback and exergetic objective values. There-
fore, the total deviation at 3.5

C denoted by d
3:5
+
C

is,
d
3:5
+
C

= d
40:6
+
C

d
18:8
+
C

(52c)
Similarly, for systems designed at optimization base tempera-
tures of 18.8

C and 40.6

C, deviations d
18:8
+
C

and d
40:6
+
C

are ob-
tained as,
d
40:6
+
C

_
p
18:8
+
C
n
p
40:6
+
C
n
_
2

18:8
+
C
n

40:6
+
C
n
_
2
_
(53a)
d
3:5
+
C

_
p
18:8
+
C
n
p
3:5
+
C
n
_
2

18:8
+
C
n

3:5
+
C
n
_
2
_
(53b)
d
18:8
+
C

= d
40:6
+
C

d
18:8
+
C

(53c)
d
18:8
+
C

_
p
40:6
+
C
n
p
18:8
+
C
n
_
2

40:6
+
C
n

18:8
+
C
n
_
2
_
(54a)
Table 4
Specications of the recuperator and the regenerative gas cycle specied by the fuzzy decision-maker at various ambient temperatures.
Parameters T
0
= 3.5

C T
0
= 18.8

C T
0
= 40.6

C
Tube arrangement Triangle(30e60) Triangle(30e60) Triangle(30e60)
Tube outside/inside diameter d
i
/d
o
(m) 0.0173/0.01905 0.0173/0.01905 0.0173/0.01905
Tube pitch ratio in the tube bundle 1.33 1.33 1.25
Tube length (m) 7.683 7.674 7.688
Number of tubes(N
t
) 6805 6572 6348
Number of bafes(N
b
) 5 7 7
Shell inside diameter D
s
(m) 2.280 2.240 2.070
Outer tube limit D
otl
(m) 2.008 1.850 1.689
Inner tube limit D
itl
(m) 1.026 1.008 0.931
Outlet temp. of the recuperator (

C) 397 412 427


Effectiveness 0.5935 0.5930 0.5925
Recuperator cost($) 470040 457240 445180
Payback time(years) 2.23 2.73 2.84
Exergetic efciency (%) 28.62 26.64 24.65
Improvement in the exergetic efciency
compared to the simple cycle (%)
2.26 1.81 1.72
Table 5
Specications of the recuperator and the regenerative gas cycle specied by the LINMAP decision-maker at various ambient temperatures.
Parameters T
0
= 3.5

C T
0
= 18.8

C T
0
= 40.6

C
Tube arrangement Triangle(30e60) Triangle(30e60) Triangle(30e60)
Tube outside/inside diameter d
i
/d
o
(m) 0.0173/0.01905 0.0173/0.01905 0.0173/0.01905
Tube pitch ratio in the tube bundle 1.33 1.33 1.33
Tube length(m) 7.665 7.659 7.702
Number of tubes(N
t
) 6118 5975 5642
Number of bafes(N
b
) 7 7 7
Shell inside diameter D
s
(m) 2.160 2.140 2.080
Outer tube limit D
otl
(m) 1.821 1.842 1.689
Inner tube limit D
itl
(m) 0.972 0.963 0.936
Outlet temp. of the recuperator (

C) 395 411 425


Effectiveness 0.5880 0.5873 0.5828
Recuperator cost($) 431990 423850 405330
Payback time(years) 2.13 2.63 2.70
Exergetic efciency (%) 28.52 26.58 24.57
Improvement in the exergetic efciency
compared to the simple cycle (%)
2.16 1.75 1.64
Table 6
Specications of the recuperator and the regenerative gas cycle specied by the TOPSIS decision-maker at various ambient temperatures.
Parameters T
0
= 3.5

C T
0
= 18.8

C T
0
= 40.6

C
Tube arrangement Triangle(30e60) Triangle(30e60) Triangle(30e60)
Tube outside/inside diameter d
i
/d
o
(m) 0.0173/0.01905 0.0173/0.01905 0.0173/0.01905
Tube pitch ratio in the tube bundle 1.33 1.33 1.33
Tube length(m) 7.667 7.659 7.724
Number of tubes(N
t
) 6109 5975 5663
Number of bafes(N
b
) 7 7 7
Shell inside diameter D
s
(m) 2.160 2.140 2.080
Outer tube limit D
otl
(m) 1.909 1.842 1.666
Inner tube limit D
itl
(m) 0.969 0.963 0.936
Outlet temp. of the recuperator (

C) 395 411 425


Effectiveness 0.5879 0.5873 0.5834
Recuperator cost($) 431520 423850 406850
Payback time (years) 2.13 2.63 2.70
Exergetic efciency (%) 28.52 26.58 24.58
Improvement in the exergetic efciency
compared to the simple cycle(%)
2.16 1.75 1.65
H. Sayyaadi, R. Mehrabipour / Energy 38 (2012) 362e375 372
d
3:5
+
C

_
p
40:6
+
C
n
p
3:5
+
C
n
_
2

40:6
+
C
n

3:5
+
C
n
_
2
_
(54b)
d
40:6
+
C

= d
3:5
+
C

d
18:8
+
C

(54c)
A solution with a minimum deviation from optimal solution of
other two ambient temperatures is desired. In this regard, our
calculation indicates,
d
3:5
+
C

= 0:001936; d
40:6
+
C

= 0:00247 and d
18:8
+
C

= 0:001583 (55)
Therefore, as is clear, optimization at the average annual
ambient temperature leads to less deviation compared to other
temperatures. Therefore, we conclude that optimization at 18.8

C
with the TOPSIS decision-maker leads to more reasonable results.
This fact can be found also by looking at results indicated on the
sixth column Table 7. This column indicates that a more efciency
improvement is obtained at the average ambient temperature by all
decision-making methods. On the other hand, as we can see from
Figs. 3aec and 4, the TOPSIS and LINMAP recommend same solu-
tion. As a summary, the nal optimal solution selected by LINMAP
and TOPSIS decision-makers at 18.8

C is selected our desired
solution for the regenerative gas cycle. It can be concluded that in
similar works on optimization of the gas cycle (which operate in
variable ambient temperature), optimization might be performed
at the average annual ambient temperature. Table 8 indicates
specications of the selected nal optimal regenerative gas cycle.
Further Figs. 5(a and b) compare the fuel consumption and exer-
getic efciency of the simple V93.1 gas cycle with improved
regenerative gas cycle in a more elucidating manner.
As is found from Table 8 and Figs. 5a, b the exergetic efciency
and fuel consumption of the modied cycle are improved 1.75% and
6.6% compared to the present Brayton V93.1 cycle. This improve-
ment requires an investment of approximately 423,850 US $ which
its cost will be paid back within 2.63 years with the current
domestic cost of the natural gas fuel in Iran. It is required to
mention that since optimization was performed based on the local
Table 7
Exergetic efciency improvement obtained by different decision-making methods at various ambient temperatures.
Decision-making
method
Optimization base
temperature (

C)
Ambient
temp. (

C)
Exergetic
efciency (%)
Exergetic efciency
improvement
Sum of efciency
improvement
in other two cases (%)
Average efciency
improvement (%)
Fuzzy 18.8 40.6 24.66 1.73 5.49 1.63
3.5 28.31 1.95
40.6 18.8 26.23 1.40 4.56
3.5 27.80 1.44
3.5 18.8 26.58 1.75 4.74
40.6 23.66 0.73
TOPSIS 18.8 40.6 24.61 1.68 5.23 1.57
3.5 28.16 1.80
40.6 18.8 26.15 1.32 4.29
3.5 27.68 1.32
3.5 18.8 26.58 1.75 4.60
40.6 23.62 0.69
LINMAP 18.8 40.6 24.61 1.68 5.23 1.61
3.5 28.16 1.80
40.6 18.8 26.32 1.49 4.13
3.5 27.36 1.00
3.5 18.8 26.12 1.29 5.14
40.6 24.62 1.69
Fig. 4. Pareto optimal frontiers at minimum, average and maximum annual temperatures indicating various decision-making methods.
H. Sayyaadi, R. Mehrabipour / Energy 38 (2012) 362e375 373
natural gas price in Iran which is very cheap in comparison to the
international market, the payback time for the optimized cycle is
relatively high. If optimization is performed with the international
fuel price which is 3.75 times higher than its corresponding value in
Iran, the payback time will be reduced to 17 months which is
reasonable value for the payback time.
8. Conclusions
A simple Brayton gas cycle namely as the Kraftwerk V93.1
Siemens gas turbine was improved by converting it into a regener-
ative gas cycle having an optimized vertical tubular recuperative
heat exchanger. It was shown that optimization process can be
employed even for existing cycles which are not in design stage.
Therefore the existing gas cycle was improved using an
optimization approach. Specications of the recuperator and its
role in the gas cycle were found in a multi-objective optimization
process using the NSGA-II algorithm. Optimization objective were
the exergetic efciency of the gas cycle and the payback time for the
recuperative investment. Multi-objective optimization of a gas
cycle as a benchmark for any energy system has been performed
with three famous decision-making methods. In this regard, the
fuzzy Bellman-Zadeh, LINMAP and TOPSIS decision-making
methods were employed for selection of a nal optimal solution
from the Pareto frontier. Further, multi-objective optimization was
performed at three ambient temperature cases including
minimum, average and maximum ambient air temperatures.
However there is no decision-making method having superiority
over other methods, It was found that in our case optimization base
temperature equal to the average ambient air temperature and the
TOPSIS decision-making leads to best results. Further it was found
that at the average ambient air temperature both the TOPSIS and
LINMAP decision-makers suggest the same nal solution for the
regenerative gas cycle. It was shown that optimization at the
average ambient temperature leads to better results than any other
systems that are optimized at other ambient reference tempera-
tures. It was discussed that in a very lowprice for the natural gas in
Iran the payback time for the optimized cycle is relatively high
whereas if the optimization performed at international for the
natural gas price, the payback time of the optimized cycle will be in
a reasonable range.
Acknowledgment
This research work has been completed by the nancial support
of Fars Regional Electric Company.
References
[1] Mortazavia A, Somersa C, Alabdulkarema A, Hwang Y, Radermachera R.
Enhancement of APCI cycle efciency with absorption chillers. Energy 2010;
35:3877e82.
[2] Al-Ibrahim AM, Varnhama A. A review of inlet air-cooling technologies for
enhancing the performance of combustion turbines in Saudi Arabia. Applied
Thermal Engineering 2010;30:1879e88.
[3] Wan K, Zhang S, Wang J, Xiao Y. Performance of humid air turbine with
exhaust gas expanded to below ambient pressure based on microturbine.
Energy Conversion and Management 2010;51:2127e33.
[4] Budzianowski WM, Miller R. Toward improvement in thermal efciency and
reduced harmful emissions of combustion process by using recirculation of
heat and mass: a review. Recent Patents on Mechanical Engineering 2009;2:
228e39.
[5] Budzianowsky WM. Thermal integration of combustion-based energy gener-
ators by heat recirculation. Rynek Energii 2010;91(6):108e15.
[6] Budzianowsky WM. A comparative framework for recirculating combustion of
gases. Archivum Combustionis 2010;30(1e2):25e36.
[7] AldushinA P, Matokowsky BJ, Volpert VA. Enhancement of Gasless combustion
Synthesis by Counterow gas Filtration. Combustion Science and Technology
1994;103:41e61.
[8] Budzianowski WM, Miller R. Superadiabatic lean catalytic combustion in
a high-pressure reactor. International Journal of Chemical Reaction Engi-
neering 2009;7. art. no. A20.
[9] Paul B, Datta A. Burner development for the reduction of NO
x
emission from
coal red electric utilities. Recent Patents on Mechanical Engineering 2008;1:
175e89.
[10] Ruixian C, Lixia J. Analysis of the recuperative gas turbine cycle with a recu-
perator located between turbines. Applied Thermal Engineering 2006;26:
89e96.
[11] Kim TS, Hwang SH. Part load performance analysis of recuperated gas turbines
considering engine conguration and operation strategy. Energy 2006;31:
260e77.
[12] Shirmohammadi M, Sayyaadi H. Multi-Objective Exergetic, Economic and
Environmental Optimization of CGAM Problem Using Genetic Algorithm.
Proceeding of 3rd International Energy, Exergy and Environment Symposium;
1e5 July 2007. Evora, Portugal.
[13] Sayyaadi H. Multi-Objective approach in Thermoenvironomic optimization of
a benchmark cogeneration system. Applied Energy 2009;86:867e79.
Table 8
Specications of the nal optimal regenerative gas cycle obtained at the average
ambient temperature of 18.8

C and TOPSIS (and LINMAP) decision-making
methods.
Parameters Value
Tube arrangement Triangle(30

)
Tube inside/outside diameter (mm) 17.3/19.05
Tube length (m) 7.659
Number of tubes 5975
Number of bafes 7
Shell inside diameter (m) 2.14
Tube pitch to tube outside diameter 1.33
Outer tube limit D
otl
(m) 1.842
Inner tube limit D
itl
(m) 0.963
Shell side heat transfer coefcient(h
s
) (W m
2
K
1
) 343.7
Tube side heat transfer coefcient(h
t
) (W m
2
K
1
) 129.38
Total heat transfer coefcient U
o
(W m
2
K
1
) 168.3
DTm(
+
C) 72.65
Tube side Pressure drop(DP
t
)(kPa) 5.02
Shell side Pressure drop(DP
s
)((kPa) 2.39
Approximate pre-heater cost($) 423850
Payback(year) 2.63
Exergy efciency improvement (%) 1.75
Fuel consumption reduction (%) 6.6
Fig. 5. Comparison of the present Brayton V93.1 gas cycle with the modied cycle in
(a) fuel consumption; (b) Exergetic efciency.
H. Sayyaadi, R. Mehrabipour / Energy 38 (2012) 362e375 374
[14] Fiaschi D, Tapinassi L. Exergy analysis of the recuperative auto thermal
reforming (R-ATR) and recuperative reforming (R-REF) power cycles with CO
2
removal. Energy 2004;29:12e5. 2003-2024.
[15] Toffolo A, Lazzaretto A. Energy, economy and environment as objectives in
multi-criteria optimization of thermal systemdesign. Energy 2004;29:1139e57.
[16] Chang CT, Hwang JR. A multi-objective programming approach to waste
minimization in the utility systems of chemical processes. Chemical Engi-
neering Science 1996;51(16):3951e65.
[17] Roosen P, Uhlenbruck S, Lucas K. Pareto optimization of a combined cycle
power system as a decision support tool for trading off investment vs. oper-
ating costs. International Journal of Thermal Sciences 2003;42(6):553e60.
[18] Valds M, Durn MD, Rovira AA. Thermo-economic optimization of combined
cycle gas turbine power plants using genetic algorithms. Applied Thermal
Engineering 2003;23(11):2169e82.
[19] Marechal F, Kalitventzeff B. Targeting the integration of multi-period utility
systems for site scale process integration. Applied Thermal Engineering 2003;
23(14):1763e84.
[20] Uhlenbruck S, Lucas K. Exergoeconomicallyeaided evolution strategy applied
to a combined cycle power plant. International Journal of Thermal Sciences
2004;43(3):289e96.
[21] Sayyaadi H, Aminian HR. Design and optimization of a Non-TEMA type tubular
recuperative heat exchanger used in a regenerative gas turbine cycle. Energy
2010;35:1647e57.
[22] Bellman R, Zadeh LA. Decision making in a fuzzy environment. Management
Sci 1970;17:141e64.
[23] Yu PL. Multiple-Criteria decision making, concepts, Techniques, and Exten-
sions. New York: Plenum Press; 1985.
[24] Olson DL. Decision Aids for selection problems. New York: Springer;
1996.
[25] Taal M, Bulatov Klemes J, Stehlik P. Cost estimation and energy price forecasts
for economic evaluation of retrot projects. Applied Thermal Engineering
2003;23:1819e35.
[26] Fonseca CM, Fleming PJ. Multiobjective optimization. In: Back T, Fogel DB,
Michalewicz Z, editors. Handbook of evolutionary Computation. Oxford
University Press; 1997.
[27] Carvalho MB, Ya Ekel P, Martins CAPS, Pereira JG. Fuzzy set-based multi-
objective allocation of resources: solution algorithms and applications.
Nonlinear Analysis 2005;63:715e24.
[28] Mazur V. Fuzzy exergoeconomic optimization of energy-transforming
systems. Applied Energy 2007;84:749e62.
H. Sayyaadi, R. Mehrabipour / Energy 38 (2012) 362e375 375

S-ar putea să vă placă și