Chemical Engineering Research and Design 1 3 1 ( 2 0 1 8 ) 29–40

Contents lists available at ScienceDirect

Chemical Engineering Research and Design

journal homepage: www.elsevier.com/locate/cherd

Simulation and optimization of a combined cycle

gas turbine power plant for part-load operation

Zuming Liu, I.A. Karimi ∗

Department of Chemical & Biomolecular Engineering, National University of Singapore, 4 Engineering Drive 4,
Singapore 117585, Singapore

a r t i c l e i n f o a b s t r a c t

Article history: Combined cycle gas turbine (CCGT) power plants must often run at part-load conditions, as
Received 3 September 2017 the electricity demand varies constantly. We present a method and necessary correlations
Received in revised form 30 for simulating the part-load operation of a typical CCGT plant in a commercial simulator
November 2017 (e.g. GateCycle). We show that assuming constant values for some equipment parameters
Accepted 4 December 2017 (e.g. efficiencies) and ignoring the operating maps of key equipment can overestimate plant
Available online 13 December 2017 performance significantly at part-loads. Furthermore, a rise in the ambient temperature
lowers the plant capacity, but increases the plant efficiency. Then, we propose a simulation-
Keywords: based optimization approach that yields an optimal operating strategy to maximize the
Power plant overall plant efficiency for any part-load. Our strategy forms a basis for evaluating the two
Simulation widely used operating policies (fuel flow control or FFC and inlet guide vane control or IGVC).
Optimization Our proposed strategy increases the plant efficiency by as much as 2.63% (absolute) over FFC
Gas turbine and 0.93% over IGVC. This work highlights the need for integrating the two cycles (gas turbine
Steam cycle and steam) to optimize the plant performance. We find that FFC seems to prioritize the gas
Part-load operation turbine and IGVC tends to prioritize the steam cycle, while our proposed strategy strikes an
optimal balance between the two.
© 2017 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

1. Introduction by their governments. Furthermore, the power demand fluctuates sig-

Natural gas is quickly replacing coal as the preferred fuel for power
generation worldwide due to its cleaner nature and lower CO2 emis-
sions (Boyce, 2012). Combined cycle gas turbine (CCGT) power plants
that use natural gas as the fuel are among the most efficient with a
thermal efficiency as high as 60% (Rao, 2012). As a result, CCGT plants
are now undergoing widespread installations. Some countries such as
Singapore produce more than 96% of their electricity from CCGT power
plants (EMA Singapore, 2017).
A CCGT power plant generates power from two cycles: A Brayton
cycle followed by a Rankine cycle. An air compressor followed by a gas
combustor and then a gas turbine (GT) are the main components of the
Brayton cycle. Multiple pumps feeding water to a heat recovery steam
generator (HRSG) that in turn supplies steam at multiple pressure levels
to steam turbines (STs) are the main components of the Rankine cycle.
Power plants including the CCGT plants are often designed with sur-
plus capacity due to the need to maintain spinning reserves mandated
nificantly with time each day. Therefore, a CCGT plant must often run
at part-loads (which means off-design conditions). This lowers its ther-
mal efficiency, wastes substantial non-renewable fossil fuels (natural
gas in this particular case), and increases CO2 emissions. Thus, simu-
lating and optimizing the performance of a CCGT plant at part-loads is
of much interest.
Much modeling work exists for simulating individual components
of a CCGT plant. Kim and Ro (1995) reported a simulation program to
evaluate the effect of configuration on the performance of a heavy-
duty GT. Zhang and Cai (2002) defined several reduced parameters
to express compressor and turbine characteristics, and used them to
derive an analytical solution for predicting the part-load performance
of a GT. Haglind and Elmegaard (2009) proposed two models to pre-
dict the part-load performance of an aero-derivative GT. One model
used the actual performance maps of the compressor and turbine,
but the other simpler model assumed turbine constants with com-
pressor maps. Irrespective, both models offered good predictions of

∗
Corresponding author.
E-mail address: cheiak@nus.edu.sg (I.A. Karimi).
https://doi.org/10.1016/j.cherd.2017.12.009
0263-8762/© 2017 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
30 Chemical Engineering Research and Design 1 3 1 ( 2 0 1 8 ) 29–40

Nomenclature
IPP Intermediate pressure pump
Symbols LP Low pressure
A Area, m2 LPP Low pressure pump
C Heat capacity, kJ/K RHT Reheater
cp Specific heat, kJ/kg/K RP Recirculation pump
c1 , c2 IGV angle correction factors SPHT Superheater
h Mass enthalpy, kJ/kg ST Steam turbine
LHV Lower heat value, kJ/kg TET Turbine exhaust temperature
m Mass flow, kg/s TIT Turbine inlet temperature
N Shaft rotational speed, rpm
NTU Number of transfer unit
P Pressure, bar mass flows and pressure profiles for the entire load range, and ther-
PR Overall pressure ratio mal efficiencies and exhaust temperatures for part-loads above 70%.
P Pressure loss, kPa Lee et al. (2011) developed a general simulation program for simple,
Q Heat duty, kW recuperative, and reheat GTs. They used a stage-stacking method for
R Gas constant the air compressor, and a stage-by-stage model for the turbine. Their
method is useful when the compressor/turbine performance maps are
T Temperature, K
not available. Song et al. (2015) combined two existing cooling mod-
U Overall heat transfer coefficient, kJ/(s m2 K)
els (Consonni, 1992; Holland and Thake, 1980) to accurately predict
W Power, MW
the cooling air flows for a GT, and analyzed their influence on the off-
design performance. Tsoutsanis et al. (2014, 2015) introduced a novel
Greek letters
map-tuning method to improve the accuracy and fidelity of GT models
˛ IGV angle for predicting the performance under steady, transient, and degraded
 Specific heat ratio conditions.
ı Expansion ratio Relative to the GTs, HRSG modeling has received much less atten-
ε Heat transfer effectiveness tion. Its off-design simulation requires estimates of overall heat
 Efficiency transfer coefficients (OHTCs). Xu et al. (2015) adopted a power law cor-
 Constant relation for this by assuming the gas side to control the heat transfer.
 Constant Kim and To (1997) and Haglind (2011) accounted for the properties
of both gas and water/steam, and used different OHTC correlations
 Specific volume
for various heating surfaces. Zhang et al. (2016b) adopted Ganapathy’s
ϕ ( − 1)/
HRSG simulation method (Ganapathy, 1990), in which both HRSG design

Cooling effectiveness
and exhaust gas parameters are considered to calculate the OHTCs at
Flow coefficient off-design conditions.
 Combustor loading The steam turbines (STs) are often modeled using the Flugel equa-
 Turbine blade cooling constant tion (Haglind, 2011) or Stodola’s method (Sanchez Fernandez et al.,
2016).
Subscripts/superscripts Plant operation/control strategies obviously impact the perfor-
a Air mance of CCGT plants significantly. Kim and Hwang (2006) evaluated
b Turbine blade three part-load operation strategies (fuel flow control (FFC), variable
c Compressor speed control (VSC), and IGV control (IGVC)) for a single-shaft recu-
ca Cooling air perative GT, and two strategies (FFC and variable area nozzle control
(VANC)) for a two-shaft recuperative GT. VSC gave the best part-load
cc Combustion chamber
performance for the former, and VANC for the latter. Haglind (2011)
cor Corrected value
analyzed the effects of variable geometry GTs on the part-load per-
d Design condition
formance of a single-pressure CCGT plant and found that the IGVs
f Fuel and VAN improved performance. Kim et al. (2003) showed that IGVC
g Flue gas increases the performance of a single-shaft combined cycle, but not a
in Inlet two-shaft configuration. Jiménez-Espadafor Aguilar et al. (2014) ana-
map Performance map lyzed eight operating strategies for a combined heat and power (CHP)
out Outlet plant based on two two-shaft GTs, and showed that IGVC offered the
Pump Feed water pump best regulation capacity. Barelli and Ottaviano (2015) proposed a novel
t Turbine/turbine stage combined cycle by adding an additional variable speed compressor
w Water/steam upstream of the GT to adjust air flow in order to improve the operational
flexibility and part-load performance.
* Critical value
The above discussion suggests that the existing work has addressed
some components of CCGT plants individually. However, the off-design
Acronyms
operations of other components such as water pumps, steam tur-
CCGT Combined cycle gas turbine plant
bines, gas combustor, and generator have received little or no attention.
ECON Economizer Furthermore, a study addressing the off-design performance of all com-
EVAP Evaporator ponents in a holistic manner does not exist. Most studies on operating
HP High pressure strategies have focused solely on the GT operation, and assumed a few
HPP High pressure pump pre-defined policies. The impact of steam cycle (SC) and its interaction
HRSG Heat recovery steam generator with the GT operation have been not studied in a holistic or integrated
IP Intermediate pressure manner.
This work aims to simulate and optimize the part-load opera-
tion of a CCGT plant while considering the off-design operation of
 Chemical Engineering Research and Design 1 3 1 ( 2 0 1 8 ) 29–40 31

Fig. 1 – Schematic of a triple-pressure reheat combined cycle power plant.

all its components in an integrated manner. Instead of implementing

these off-design models in a stand-alone system, we cleverly embed
them inside a commercial power plant simulator such as GateCycle
(General Electric, 2013). While other simulators such as Thermoflex
(Thermoflow, 2016) and Ebsilon Professional (2016) can also be used,
most simulate the operation for a given set of constant parame-
ters and ignore operating maps. For example, they all assume that
combustor pressure drop, combustor efficiency, GT efficiency, pump
efficiency, etc. are constant. This is obviously inaccurate, and our
methodology addresses this gap. Furthermore, a commercial software
can only simulate the operation for a given policy, but it cannot opti-
mize the policy. In other words, one must manually test pre-formulated
operation strategies to improve part-load operations. Therefore, we
present a comprehensive multi-variable simulation-based optimiza-
tion approach to find the best operation strategy for any given part-load.
We then compare our optimal strategy with two widely used operating
policies in practice. Fig. 2 – Relativized compressor map for the example CCGT
plant (Relative refers to a parameter divided by its design
point value).
2. Typical CCGT power plant
off-design models for all its components under the following
Fig. 1 shows the schematic of a typical CCGT power plant.
assumptions.
The pressurized ambient air from the air compressor enters
the combustor, where a gaseous fuel such as natural gas is
burnt. The combustion gas expands in the turbine to produce (1) The CCGT power plant operates at steady state;
power through a generator. Some air from one or more stages (2) The air compressor is equipped with IGVs, but the GT has
of the compressor is supplied to the turbine blades to keep no VAN.
them cool. The exhaust gas from the turbine enters an HRSG (3) The pumps, air compressor, and turbines are all adiabatic.
that comprises sets of steam generator modules (each module (4) There is no water leakage or blow-down in the HRSG.
has one or more economizers, evaporators, superheaters, and
desuperheaters). Each steam generator module recovers the 3. Mathematical model
waste heat from the exhaust gas of the GT to generate steam
at a certain pressure level (typically LP, IP, and HP). A steam We now develop models that accurately predict the off-design
turbine with several sections (typically LP, IP, and HP) on a sin- performance of various CCGT plant components from given
gle shaft uses the steam at various pressure levels from the design parameters and conditions.
HRSG to generate electricity. The exhaust gas from the HRSG
is finally vented to the atmosphere as stack gas. 3.1. Air compressor
A series of pumps supplies feed water to various econo-
mizers. The exhaust steam from the final section of the ST Its performance at the design conditions is represented by
is condensed in a condenser and, the condensate is pumped air mass flow (min,d ), pressure ratio (PRd ), and isentropic effi-
back to LP economizer. In this work, we simulate and optimize ciency (c,d ). Under off-design conditions, its operation follows
the part-load performance of a CCGT plant by integrating the a performance map such as the one shown in Fig. 2. This
32 Chemical Engineering Research and Design 1 3 1 ( 2 0 1 8 ) 29–40

map describes the relationships among the following non- The corresponding off-design efficiency can be estimated
dimensional variables: by a semi-empirical formula (Lu and Lin, 1996; Zhang et al.,
 2016a):
Corrected mass flow : mcor = min Tin /Pin (1a) 

   
ϕ ϕ
t N Tin,d ıd − 1 N Tin,d ıd − 1
Relative corrected mass flow : mcor /mcor,d (1b) =  − ( − 1)
t,d Nd Tin ıϕ − 1 Nd Tin ıϕ − 1
(9)
Relative pressure ratio : (PR − 1) / (PRd − 1) (1c)

where  is a known constant.

Relative isentropic efficiency : c /c,d (1d)
      3.4. GT blade cooling
Relative corrected speed : N/ Tin / Nd / Tin,d (1e)
As the turbine blade rows are exposed to the hot flue gas,
A modern air compressor is usually equipped with variable cooling air is supplied to prevent overheating and maintain
inlet guide vanes (IGVs) to regulate its airflow. Its perfor- lifetime. This cooling air flow to each turbine stage changes
mance map assumes a fully open IGV (IGV angle ˛ = 0), so during the off-design operation, and can be corrected as fol-
the mcor,map and c,map correspond to ˛ = 0. Then, the cor- lows (Palmer and Erbes, 1994):
rected mass flow and isentropic efficiency for a partially open
 0.5
IGV can be computed as follows (Kurzke, 2015): Pca Tca,d
mca = mca,d (10)
Pca,d Tca
mcor = mcor,map (1 − c1 ˛) (2)
  The blade cooling performance can be predicted by a semi-
c = c,map 1 − c2 ˛2 (3) empirical cooling effectiveness relation as follows (Kim and
Ro, 1995):
where, c1 and c2 are the vane angle correction factors.
mca · cp,ca

= (11)
3.2. Combustion chamber mg · cp,g
∞ −

The combustion chamber operation is simulated using the where,

∞ is the asymptotic cooling effectiveness,  is a
energy balance: known constant reflecting the cooling technology level, and

is the blade cooling effectiveness given as follows.
ma ha + cc mf LHV + mf hf = mg hg (4)
Tg − Tb

= (12)
Tg − Tca
The heat addition and flow friction cause a pressure drop
in the combustion chamber. The two factors are known as hot
Eqs. (10)–(12) give us the blade metal temperature as fol-
and cold pressure losses, respectively. The hot pressure loss
lows.
is usually negligible, but the cold pressure loss varies with the
 
corrected combustor inlet flow as follows (Walsh and Fletcher,   (mca · cp,ca ·
∞ ) /  · mg · cp,g
2004). Tb = Tg − Tg − Tca   (13)
1 + (mca · cp,ca ) /  · mg · cp,g
   
2
Pcc min Tin min Tin 3.5. Heat recovery steam generator (HRSG)
= / (5)
Pcc,d Pin Pin
d
The HRSG is modeled as a series of heat exchangers. The
The corresponding combustor efficiency can be correlated effectiveness-number of transfer unit (NTU) method can
as follows (Kurzke, 2015), where  is the combustor loading. predict their off-design performance using the overall heat
transfer coefficient under off-design conditions given by the
min following:
= (6)
P1.8 exp (Tin /300)
in
0.8
  1.6 U mg
= (14)
(1 − cc ) = (1 − cc,d ) (7) Ud mg,d
d

Then, the effectiveness and duty of each heat exchanger

3.3. Gas turbine (GT)
during off-design operation can be determined using the fol-
lowing.
The turbine usually has multiple stages. We assume it to
always operate at the choking condition, which is char-
NTU = UA/Cmin (15a)
acterized by a constant swallowing capacity. Under this
assumption, the off-design operation of a turbine stage is  
Cmax = max (mcp )w , (mcp )g (15b)
described by the following relationship (Palmer et al., 1993;
Reynolds and Perkins, 1977).
1 − exp [−NTU (1 − Cmin /Cmax )]
 ε= (15c)
  2  +1
1 − Cmin /Cmax exp [−NTU (1 − Cmin /Cmax )]
min Tin  −1
= constant, where  = (8)  
Ain Pin Rg  +1 Q = εQmax = εCmin Tg,in − Tw,in (16)
 Chemical Engineering Research and Design 1 3 1 ( 2 0 1 8 ) 29–40 33

Fig. 4 – Feed water pump curves for the example CCGT

Fig. 3 – Steam turbine efficiency curve for the example plant (Relative refers to a parameter divided by its design
CCGT plant (Relative refers to a parameter divided by its point value).
design point value).

Then, the outlet temperatures of the gas, water, and steam

flows can be computed from the energy balance.
The pressure drops for various streams in each heat
exchanger under off-design conditions can be computed as
follows (Erbes and Gay, 1989):

P
 m 1.84  T   P −1
• Gas-side pressure drop: =
Pd md Td Pd
(17)

P
 m 1.98
• Water-side pressure drop: = (18)
Pd md

P
 m 1.98   −1 Fig. 5 – Generator efficiency curve for the example CCGT
• Stream-side pressure drop: = (19) plant (Relative refers to a parameter divided by its design
Pd md d
point value).
3.6. Steam turbine (ST)
its performance curve such as the one in Fig. 4. Its power
ST is modeled as a series of sections. A typical ST has three consumption is given by (Cengel, 2007):
sections (HP, IP and LP). A section-by-section method is com-
monly used for its simulation. The off-design characteristics of 1
WPump = mw,in w,in (Pw,out − Pw,in ) (21)
a section (HP, IP or LP) are described by the following modified Pump
Stodola equation (Erbes, 1986):
3.8. Generator
  r − r ∗ 2
min
 = 1− (20a) The generator efficiency varies with its load and can be
Pin /in 1 − r∗
obtained from an efficiency curve such as the one shown in
Fig. 5.
 2 /(−1)
This completes our correlations for modeling the part-load
r∗ = (20b)
 +1 operations of the key components in a CCGT plant. We imple-
mented these in a commercial simulator GateCycle (General
Eq. (20a) is first used to compute the flow coefficient at Electric, 2013). While GateCycle has some in-built compres-
the design condition for each section. Then, is used back in sor maps for ready use, an arbitrary map must be supplied
Eq. (20a) to predict the off-design operation. as a Bivariate Table. GateCycle does not have maps for steam
The isentropic efficiency of a section under off-design con- turbines, water pumps, and electricity generator. Hence, we
ditions is given by the ST efficiency curve (Ebsilon Professional, supplied these as Univariate Tables. GateCycle accepts only
2016) such as the one shown in Fig. 3. one constant value each for combustion chamber pressure
drop, and combustion chamber efficiency. Since these param-
3.7. Water pumps eters must change under part-load operations, we used Set
macros in GateCycle to compute their values as functions of
A water pump supplies feed water to a steam circuit (HP, IP, part-load. To demonstrate the utility of our model, we now
or LP). Its operation under off-design conditions is given by present a case study on an example CCGT plant.
34 Chemical Engineering Research and Design 1 3 1 ( 2 0 1 8 ) 29–40

in Eqs. (5) and (6), and  = 2.083 (Lu and Lin, 1996) in Eq. (12).
Table 1 – Design parameters and performance of the
example CCGT plant. Fig. 3 gives the relative efficiency curves for the HP/IP/LP steam
turbines, Fig. 4 gives the relative operating maps for the water
Variable/parameter Value
pumps, and Fig. 5 gives the relative efficiency curve for the
Air compressor electricity generator.
Air inlet pressure (kPa) 101.3 We use GateCycle (General Electric, 2013) to study the
Air inlet temperature (◦ C) 15.0
plant’s performance at part-loads ranging from 40 to 100%
Relative humidity of air (%) 60.0
under two operating policies (FFC and IGVC). In FFC, the
Inlet pressure loss (%) 0.5
Airflow (kg s−1 ) 635.0 fuel flow is reduced to achieve the desired part-load. In
Number of stages 18 IGVC, IGVs and fuel flow are changed simultaneously to
Pressure ratio 15.4 achieve the desired part-load, while maintaining the tur-
Isentropic efficiency (%) 88.0 bine exhaust temperature (TET) at its design value. In
Fuel combustor contrast to these known control policies, we propose a
Fuel flow (kg s−1 ) 14.74 novel multi-variable optimization strategy to maximize the
Exit temperature (◦ C) 1405.0 plant’s thermal efficiency at any given part-load. However,
Combustor efficiency (%) 99.5 first, we use our model to shade some light on the errors
Pressure loss (%) 3.5
introduced by assuming some parameters as constant in a
Gas turbine (GT) simulation tool such as GateCycle (General Electric, 2013)
Number of stages 3 for studying the part-load operation of this plant under
Stage efficiency (%) 87.2 FFC and IGVC. For the same policies, we also study the
Exit pressure loss (%) 0.5
impact of ambient parameters (e.g. ambient humidity and
Exit temperature (◦ C) 615.0
temperature) on the full-range performance of this CCGT
Heat recovery steam generator (HRSG) plant.
HP/IP/LP steam temperatures (◦ C) 510.0/297.0/295.0
HP/IP/LP pinch point temperatures (◦ C) 10.0/10.0/10.0
HP/IP/LP approach point temperatures (◦ C) 18.0/20.0/34.7 4.1. Errors in performance estimation from neglecting
Pressure Loss on gas/water/steam sides (%) 1.0/5.0/3.0 off-design corrections
Overall heat transfer coefficient (W m−2 ◦ C−1 ) 45.43
Stack gas temperature (◦ C) 99.7 We compare two approaches (A and B) for estimating the
Steam turbine (ST) plant’s part-load performance. In A, we simply input the
HP/IP/LP section inlet pressures (bar) 98.8/24.0/4.0 design parameters from Tables 1 and 2 in GateCycle (General
HP/IP/LP section inlet temperatures (◦ C) 565.0/565.0/312.5 Electric, 2013), and keep them the same for all part-loads. Since
HP/IP/LP section efficiencies (%) 87.0/91.0/89.0 GateCycle (General Electric, 2013) takes only single values for
Condenser combustor pressure drop, combustor efficiency, GT efficiency,
Pressure (kPa) 7.4 ST efficiency, pump efficiency, and generator efficiency, we set
Heat transfer coefficient (kW m−2 ◦ C−1 ) 2.39 them at their design values from Tables 1 and 2. In B, how-
Cooling water temperature (◦ C) 25.0
ever, we implement our off-design corrections (Eqs. (5–7) and
Generator (9)) for these parameters in GateCycle (General Electric, 2013),
Speed of rotation (RPM) 3000 thus their values vary precisely with part-load. In addition, we
Mechanical efficiency (%) 99.0 use the following in B.
Generator efficiency (%) 98.5

CCGT performance
• Full-range efficiency curves for the STs from Fig. 3.
GT power (MW) 257.2
GT efficiency (%) 36.78
• Full operating maps for the water pumps from Fig. 4 in place
ST power (MW) 136.4 of the single efficiency assumed in A.
Total power (MW) 393.6 • Efficiency curve for the electricity generator from Fig. 5.
Combined cycle efficiency (%) 56.29

4. Case study
Fig. 1 shows a 393.6 MW CCGT plant with a triple-pressure
reheat HRSG. Tables 1 and 2 present its design parameters and
performance. Fig. 2 gives the map (Palmer et al., 1993) for its
air compressor. We take c1 = 0.01 and c2 = 0.0001 (Kurzke, 2015)
Fig. 6 (for FFC) and Fig. 7 (for IGVC) show the plant per-
formance at 40–100% part-loads in terms of GT efficiency
(GTE), steam cycle efficiency (SCE), and combined cycle effi-
ciency (CCE). Approach B should obviously reflect the true
plant performance at each part-load. Figs. 6 and 7 show that
the plant efficiencies are lower at part-load conditions, and
they decrease as the plant load decreases. For FFC (IGVC), A
overestimates GTE and CCE by as much as 5.0% (3.0%) and

Table 2 – Cooling air parameters for the gas turbine.

Stage Blade Metal temperature (◦ C) Cooling air bleeding stage Cooling air ratioa (%)
∞

1 Nozzle 850.0 18 7.9 0.95

1 Rotor 850.0 16 3.3 0.95
2 Nozzle 800.0 13 3.0 0.90
2 Rotor 800.0 16 2.2 0.90
3 Nozzle 750.0 9 1.8 0.85

a
Percent of the air flow.
 Chemical Engineering Research and Design 1 3 1 ( 2 0 1 8 ) 29–40 35

Fig. 6 – Plant performance predictions from approaches A and B under FFC.

Fig. 7 – Plant performance predictions from approaches A and B under IGVC.

2.7% (2.6%) respectively, and underestimates SCE by as much corrections overestimates the overall plant performance. This
as 2.0% (0.35%). Note these are absolute percentages and not highlights an important point regarding the operation of CCGT
relative. Thus, neglecting the operating maps and off-design plants at part-loads.
36 Chemical Engineering Research and Design 1 3 1 ( 2 0 1 8 ) 29–40

Fig. 8 – Effect of the air humidity on CCGT plant

performance under different ambient temperatures.

Under FFC (IGVC), the overestimation of GTE by A results

in a lower TET (turbine exhaust flow (TEF)), and thus
SCE is underestimated. However, the overestimation of GTE
outweighs the underestimation of SCE, resulting in an overes-
timation of CCE for both FFC and IGVC by A.

4.2. Effects of the ambient conditions on plant

performance

Fig. 8 shows the effect of ambient humidity on the plant per-

formance at air temperatures of 5 ◦ C, 20 ◦ C, and 35 ◦ C. The
effect is negligible, because moisture like air is also a gas, and
thus behaves like air except for different physical properties.
Since the moisture content is not very high, the overall phys-
ical properties of the ambient air remain largely unaffected.
Thus, the impact on the performance is not noticeable.
Fig. 9 for FFC and Fig. 10 for IGVC show the effect of the
ambient temperature on GTE, SCE and CCE. We see that the
plant is derated at higher ambient temperatures. In other
words, its power output decreases from 106.6% of its rated
capacity at 5 ◦ C to 86.8% at 35 ◦ C. Thus, the ambient tem-
perature has a significant impact on the capacity. However,
both SCE and CCE increase with the ambient temperature.
This is because TET increases with the ambient temperature,
which means more heat to SC, greater waste heat recovery,
and better combined cycle performance. Interestingly, GTE of
FFC decreases, while that of IGVC increases slightly at low
loads, as the ambient temperature increases.
Fig. 9 – Effect of the ambient temperature on plant
performance under FFC.
4.3. Optimization for maximum CCE
While FFC and IGVC are widely used in CCGT plants, we are
unaware of any benchmarks for their effectiveness. While a
few studies have shown that IGVC is better than FFC, it is not
established if IGVC yields the maximum CCE. To obtain the
best operating strategy, one must optimize all the degrees of
freedom in a CCGT plant and consider both cycles (Brayton and
Rankine) in an integrated manner. In other words, a rigorous
multivariable optimization strategy is needed to guarantee the
maximum CCE. We propose such a methodology based on the
Particle Swarm Optimization (PSO) algorithm (Kennedy and
Eberhart, 1995). Our strategy uses IGV angle (˛), fuel flow (mf ),
cooling airflow to turbine stage t (mca,t ), and four water flows
[desuperheaters (mDS1 , mDS2 ), recirculation (mRP ), and bypass
 Chemical Engineering Research and Design 1 3 1 ( 2 0 1 8 ) 29–40
Fig. 10 – Effect of the ambient temperature on plant
performance under IGVC.
(mbp )] as in Fig. 1] in HRSG as the optimization variables to
maximize CCE at a given part-load. Maximizing CCE is same
as minimizing the fuel flow for a given part-loadb.
Minimize mf (
˛, mca,t , mDS1 , mDS2 , mRP , mbp ) (22) maximum iterations. It takes about 8000–9000 CPU s on a Dell
We enforce the following operational constraints during
optimization:
• Each turbine blade metal temperature must not exceed
its design metal temperature:
Tb ≤ Tb,d
• HP steam and reheat steam temperatures must not
exceed their design values:
T(HP Steam) ≤ 565.0 ◦ C (24a)
37
Fig. 11 – Simulation-based optimization using PSO
algorithm.
T(Reheat Steam) ≤ 565.0 ◦ C (24b)
• Stack gas temperature must exceed a lower limit to pre-
vent corrosion:
◦
T(Stack gas) ≥ 99.7 C (25)
We used a simulation-based optimization approach (see
Fig. 11) to optimize the plant performance for any given part-
load. The plant is simulated in GateCycle (General Electric,
2013), and the PSO algorithm is implemented in Matlab
(MathWorks, 2016). GateCycle is interfaced with Matlab via
Cyclelink. The PSO particles are generated in Matlab, and then
sent to GateCycle for simulations. If the simulation converges,
then its results are sent back to Matlab for evaluating the
objective function along with the penalties for constraint vio-
lations. The PSO algorithm terminates, when the termination
condition is met. Otherwise, the particle velocity and positions
are updated, and the PSO algorithm continues. In the PSO algo-
rithm, we set 0.9 for inertia weight, 2.0 for social parameter,
2.0 for cognitive parameter, 20 for particle size, and 200 for
®
OPTIPLEX 9010 with 3.40 GHz Intel CoreTM i7-3770 CPU and
16 GB RAM to get the optimal strategy for a given part-load.
Figs. 12–16 show how the plant variables and performance
change with part-load under FFC, IGVC, and our optimal strat-
egy. For these, Tables 1 and 2 give the base values of these
variables at 100% load. The change (%) in a variable at each
(23) part-load is computed with respect to its base value in Tables
1 or 2. In Figs. 12–14, temperature normalization is done in
Kelvin instead of ◦ C. In Figs. 15 and 16, the various efficiencies
are absolute and not normalized values.
Fig. 12 shows the results for FFC. Because FFC uses only the
fuel flow to control the power output, the turbine inlet temper-
38 Chemical Engineering Research and Design 1 3 1 ( 2 0 1 8 ) 29–40
Fig. 12 – Plant operating variables versus plant load under
FFC.
Fig. 13 – Plant operating variables versus plant load under
IGVC.
ature (TIT) drops by 26.9%, as the part-load reduces from 100%
to 40%. The compressor operating point moves downward
along the constant speed line, so the pressure ratio reduces
by 13.3%. However, the airflow increases by only 1.0%. This
is why TIT decreases. The pressure ratio affects the cooling
airflows according to Eq. (10), hence the total cooling airflow
(summed over all GT stages) reduces by 10.0%. Both TET and
Fig. 14 – Plant operating variables versus plant load for the optimal strategy (The IGV opening (100 − ˛) profile is the same
as the airflow.).
blade metal temperature drop quickly with TIT. As a result, the
GT power reduces by 62.8%, and GTE by 11.3%. The lower TET
reduces the waste heat recovery in SC, hence the ST power
reduces by 54.6%, and SCE by 8.8%. Overall, CCE under FFC
decreases by 14.3% from 56.3% at 100% load to 42.0% at 40%
load. In other words, part-load operation has a huge impact
on the plant performance under FFC.
Recall that IGVC (Fig. 13) changes both the IGV angle and
fuel flow simultaneously to achieve the desired power, while
maintaining TET constant. The compressor operating point
moves nearly parallel to the surge line, reducing the airflow
by 33.4%, and pressure ratio by 38.1%. The lower air pressures
reduce the total cooling airflow by 33.6%. The simultaneous
reductions in the fuel flow and airflow retard the drop in TIT.
Thus, TIT drops by 12.2% as compared to 26.9% for FFC. In
other words, GT and HRSG run much hotter in IGVC than FFC.
Since TET remains constant, the drop in TIT reduces the GT
power by 72.9%, and GTE by 17.5%. However, the higher TET
greatly improves the SC performance compared to FFC. Hence,
the ST power reduces by 35.6% versus 54.6% for FFC, and SCE
only slightly by 0.6%. Although GTE for IGVC is lower than
that for FFC, the better use of SC gives a higher CCE for IGVC,
which reduces by only 12.6% from 56.3% to 43.7%. This clearly
suggests that IGVC is more efficient than FFC.
In contrast to FFC and IGVC, the optimal strategy (Fig. 14)
manipulates several more variables, hence gives the best CCE.
The pressure ratio reduces by 30.2%, and total cooling airflow
decreases by 81.4%. The optimizer minimizes the cooling air-
flows to keep GT as hot as allowable. Compared to FCC and
IGVC, the optimal strategy uses 47.8–71.2% lower total cool-
ing airflow. The optimizer tends to maximize the blade metal
temperature. TIT drops by 17.2%, more than IGVC, but less
than FFC. With TET reducing by 4.0% as well, GT has lower
average temperatures than IGVC. The lower TET avoids the
need for desuperheater flows, so they are all zero. The opti-
mizer keeps the stack gas temperature at its minimum to
ensure maximum heat recovery by changing the recirculation
and bypass flows. Under the optimal strategy, the GT power
and GTE reduce by 70.0% and 15.0% respectively, while the SC
power and SCE decrease by 41.0% and 1.6% respectively. How-
ever, CCE is higher than FFC and IGVC, and reduces by 11.7%
from 56.3% to 44.6%, which is the lowest drop among the three
operating policies.
 Chemical Engineering Research and Design 1 3 1 ( 2 0 1 8 ) 29–40
Fig. 15 – GT power, GTE, SC power, and SCE versus plant load under FFC, IGVC, and optimal strategy.
Fig. 16 – CCE versus plant load under FFC, IGVC, and
optimal strategy.
Comparing FFC, IGVC, and our optimal strategy, we infer
the following.
1. The widely used IGVC is not the best operating policy for a
CCGT plant.
2. Our optimal strategy gives the best part-load performance.
FFC gives the worst performance. IGVC is slightly worse
than our optimal policy.
3. CCE for the optimal strategy is at most 2.63% higher than
FFC and at most 0.93% higher than IGVC.
4. FFC seems to prioritize GTE, IGVC seems to prioritize SCE,
while the optimal strategy makes the best trade-off. In
other words, it gives the best compromise between the two
cycles (Brayton versus Rankine).
39
5. Gas temperatures in GT are the lowest for FCC, the high-
est for IGVC, and somewhere in between for the optimal
strategy.
5. Conclusion
Power plants must often run at partial capacities, thus sim-
ulating and optimizing their operations at part-loads are
important. In this work, we presented a method and appropri-
ate correlations for simulating a CCGT plant under part-load
conditions within a commercial simulator such as GateCycle.
The application of our method to study an example CCGT
plant reveals interesting observations. First, the plant per-
formance is overestimated, if the operations of key plant
components are not corrected properly for off-design condi-
tions. In fact, the overestimation increases, as the plant load
decreases. Second, as the ambient temperature increases, the
plant efficiency increases, but the plant capacity decreases.
This offers an insight to improve the efficiency via air heating
using waste heat.
In contrast to the two widely used simplistic policies (FFC
and IGVC) for operating a CCGT plant, we proposed a rigor-
ous multi-variable simulation-based optimization approach
to obtain an operating strategy that maximizes the overall
plant efficiency. Applying this strategy to the example CCGT
plant revealed the following. First, FFC prioritizes GTE, and
gives the worst plant performance. IGVC prioritizes SCE, and
gives much better performance than FFC. However, it is not the
best operation strategy for a CCGT plant. Our optimal strat-
egy achieves the best trade-off between the GT and SC cycles,
and gives the best plant performance. It can increase CCE by
as much as by 2.63% and 0.93% compared to FFC and IGVC
respectively. Our optimization approach thus provides a basis
for evaluating various operating policies.
40 Chemical Engineering Research and Design 1 3 1 ( 2 0 1 8 ) 29–40

Acknowledgements Kim, T.S., Ro, S.T., 1995. Comparative Evaluation of the Effect of
Turbine Configuration on the Performance of Heavy-duty Gas
Turbines, ASME paper 95-GT-334.
Zuming Liu acknowledges ACTSYS Process Management
Kim, T.S., To, S.T., 1997. The effect of gas turbine coolant
Consultancy Company, Singapore, for hosting his industrial modulation on the part load performance of combined cycle
internship under a ring-fenced Graduate Research Scholar- plants — part 2: combined cycle plant. Proc. Inst. Mech. Eng.
ship from the National University of Singapore. The authors A: J Power Energy 211, 453–459.
thank Mr. Norman Lee, MD of ACTSYS for inspiring them to Kurzke, J., 2015. GasTurb 12 — Design and off-design performance
work on gas turbine modeling. They further thank Mr Nor- of gas turbines. http://www.gasturb.de/.
man Lee, Dr Yu Liu, and Mr Weiping Zhang of ACTSYS for Lee, J.J., Kang, D.W., Kim, T.S., 2011. Development of a gas turbine
performance analysis program and its application. Energy 36,
several enlightening discussions and preliminary information
5274–5285.
on the GT operation in a CCGT power plant. The authors also Lu, S., Lin, R., 1996. Gas turbine steady-state design and off-design
acknowledge the use of GateCycle under academic licenses. characteristic general model. Inst. Eng. Thermophys., 404–407.
MathWorks, 2016. https://www.mathworks.com/.
References Palmer, C.A., Erbes, M.R., 1994. Simulation methods used to
analyze the performance of the GE PG6541B gas turbines
utilizing low heating value fuels, ASME IGTI Cogen-Turbo.
Barelli, L., Ottaviano, A., 2015. Supercharged gas turbine
Palmer, C.A., Erbes, M.R., Pechtl, P.A., 1993. GateCycle
combined cycle: an improvement in plant flexibility and
performance analysis of the LM2500 gas turbines utilizing low
efficiency. Energy 81, 615–626.
heating value fuels, ASME Cogen Turbo Power, ASME IGTI, 8,
Boyce, M.P., 2012. Combined Cycle Systems for Near-Zero
69.
Emission Power Generation, Combined cycle power plants, 1st
Rao, A.D., 2012. Combined Cycle Systems for Near-Zero Emission
ed.
Power Generation, Natural gas-fired combined cycle (NGCC)
Cengel, Y.A., 2007. Thermodynamics: An Engineering Approach,
systems.
6th ed. McGraw-Hill, New York (London).
Reynolds, W.C., Perkins, H.C., 1977. Engineering
Consonni, S., 1992. Performance Prediction of Gas/Steam Cycles
Thermodynamics. McGraw-Hill.
for Power Generation (Ph.D. thesis). Princeton University.
Sanchez Fernandez, E., Sanchez del Rio, M., Chalmers, H.,
Ebsilon Professional, 2016. http://www.sofbid.com.
Khakharia, P., Goetheer, E.L.V., Gibbins, J., Lucquiaud, M., 2016.
EMA Singapore, 2017. https://www.ema.gov.sg/Statistics.aspx.
Operational flexibility options in power plants with integrated
Erbes, M.R., 1986. Phased Construction of Integrated Coal
post-combustion capture. Int. J. Greenh. Gas Control 48,
Gasification Combined-Cycle Power Plants (Ph.D.
275–289.
dissertation). Stanford University, Stanford, California.
Song, Y., Gu, C.-w., Ji, X.-x., 2015. Development and validation of a
Erbes, M.R., Gay, R.R., 1989. Gate/cycle predictions of the
full-range performance analysis model for a three-spool gas
off-design performance of combined-cycle power plants.
turbine with turbine cooling. Energy 89, 545–557.
ASME Winter Annual Meeting.
Thermoflow, 2016. https://www.thermoflow.com/.
Ganapathy, 1990. Simplify heat recovery steam generator
Tsoutsanis, E., Meskin, N., Benammar, M., Khorasani, K., 2014. A
evaluation. Hydrocarb. Process. 77.
component map tuning method for performance prediction
General Electric, 2013. GateCycle. Version 6.1.3. General Electric
and diagnostics of gas turbine compressors. Appl. Energy 135,
Company.
572–585.
Haglind, F., 2011. Variable geometry gas turbines for improving
Tsoutsanis, E., Meskin, N., Benammar, M., Khorasani, K., 2015.
the part-load performance of marine combined cycles —
Transient gas turbine performance diagnostics through
combined cycle performance. Appl. Therm. Eng. 31,
nonlinear adaptation of compressor and turbine maps. J. Eng.
467–476.
Gas Turbines Power 137, 091201.
Haglind, F., Elmegaard, B., 2009. Methodologies for predicting the
Walsh, P.P., Fletcher, P., 2004. Gas Turbine Performance. John
part-load performance of aero-derivative gas turbines. Energy
Wiley & Sons.
34, 1484–1492.
Xu, Y.-j., Zhang, S.-j., Chi, J.-l., Xiao, Y.-h., 2015. Steady-state
Holland, M., Thake, T., 1980. Rotor blade cooling in high pressure
off-design thermodynamic performance analysis of a SCCP
turbines. J. Aircr. 17, 412–418.
system. Appl. Therm. Eng. 90, 221–231.
Jiménez-Espadafor Aguilar, F., Quintero, R.R., Trujillo, E.C., García,
Zhang, G., Zheng, J., Xie, A., Yang, Y., Liu, W., 2016a.
M.T., 2014. Analysis of regulation methods of a combined heat
Thermodynamic analysis of combined cycle under
and power plant based on gas turbines. Energy 72, 574–589.
design/off-design conditions for its efficient design and
Kennedy, J., Eberhart, R., 1995. Particle swarm optimization.
operation. Energy Convers. Manage. 126, 76–88.
Proceeding of IEEE International Conference on Nerual
Zhang, G., Zheng, J., Yang, Y., Liu, W., 2016b. Thermodynamic
Neworks.
performance simulation and concise formulas for
Kim, J.H., Kim, T.S., Sohn, J.L., Ro, S.T., 2003. Comparative analysis
triple-pressure reheat HRSG of gas–steam combined cycle
of off-design performance characteristics of single and
under off-design condition. Energy Convers. Manage. 122,
two-shaft industrial gas turbines. J. Eng. Gas Turbines Power
372–385.
125, 954.
Zhang, N., Cai, R., 2002. Analytical solutions and typical
Kim, T., Hwang, S., 2006. Part load performance analysis of
characteristics of part-load performances of single shaft gas
recuperated gas turbines considering engine configuration
turbine and its cogeneration. Energy Convers. Manage. 43,
and operation strategy. Energy 31, 260–277.
1323–2337.