Académique Documents
Professionnel Documents
Culture Documents
ABSTRACT
In cooperation with Rolls-Royce and Lund University, a
combined cycle power plant was modelled, simulated and optimised in the new software framework, the GTPOM tool, developed through the recent EC FP5 programme. The goal was
not only to calibrate its component models but also to show
how this new tool can be used.
This paper presents the gas turbine and combined cycle
whole plant modelling and its life cycle cost optimisation
within the GTPOM tool. In addition, this paper discusses the
software tool bearing in mind that the existence of other commercially available software packages for the purpose.
The other thermal engineering software package referred in
this project is a product of Thermoflow Inc and is a combination of three software modules, GTPRO, GTMASTER and
PEACE. In this project, data from this package was used to
calibrate the CCGT plant model in the GTPOM tool.
Rolls-Royce Industrial Trent 50 gas turbine was modelled in
the GTPOM tool with the abilities to calculate both designpoint and part-load performance for given ambient temperature
and exhaust loss.
The CCGT power plant with a 2-on-1 configuration was
modelled in the GTPOM tool. The component model library
developed in the GTPOM project was used, however the component models were modified and improved through calibrations with the reference GTPRO data.
The life cycle cost and through life economic parameters of
the CCGT whole plant in GTPOM have also been analysed,
and then this whole plant model was used as the non-optimised
benchmark model for a study in which the model was optimised in the GTPOM tool using a GA optimiser in order to
maximise the Internal Rate of Return (IRR).
NOMENCLATURE
Abbreviations
CCGT
Combined Cycle Gas Turbine
DLE
Dry Low Emission
EC
European Commission
GA
Genetic algorithm
GT
Gas Turbine
GTPOM Gas Turbine Plant Optimisation
HP
High Pressure
HRSG
Heat Recovery Steam Generator
IP
Intermediate Pressure
IRR
Internal Rate of Return
LMTD Logarithmic Mean Temperature Difference
LP
Low Pressure
PEACE Plant Engineering and Construction Estimator
ST
Steam turbine
Designations
C
Correction factor for number of tube rows
h
Specific enthalpy
mf
Mass Flux
Pr
Prandtl number
Re
Reynolds number
x1,2,3,4
Steam turbine coefficients
Y
ST output
Genetic algorithm (GA) is used as the optimisation technique. The principle of GA was developed by JH Holland and
is based on the concept of survival of the fittest [i]. Reference
[ii] gives a good overview of this method and it also gives an
example in the heat and power plant field. The variable to be
optimised i.e. efficiency or internal rate of return is called the
objective function while the variables or parameters to be intentionally varied i.e. approach temperatures, pressures etc. are
called decision variables. First, multiple calculations for sampling purposes happen with randomly selected values of decision variables within allowable ranges specified, and each value
of the decision variables is coded into a binary string called
phenotype. Then the selection of the phenotypes as the natures
of individuals starts by finding the fitness of each individual
according to the objective function e.g. maximizing the efficiency. Only the best individuals will survive to breed. Different combinations of two parents (phenotypes) create offspring
and the new generation is being evaluated. Also a small population goes through a mutation, a small change in the binary code
to generate solutions that were not included in the starting
population [iii].
Using GA a complex mathematical plant model can be optimised without requiring derivative information, however according to [iv] the GAs are computationally expensive and
since GA is based on a random process the efficiency of solving a given problem is impossible to predict. However, the
beauty of GA in optimisation problems is its robustness when a
system with multiple decision variables has non-linear behaviour and/or multiple optima. Compared to full/pure random
search, in addition, it is relatively less time consuming to seek a
global optimum.
CCGT PLANT MODELLING AND OPTIMISATION
In this study, the Rolls-Royce Trent 50 was modelled in the
GTPOM tool as a preparation for the combined cycle modelling. The Trent 50 is an industrial gas turbine, with Dry Low
Emission (DLE), derived from the aero engine Trent 800 that
powers some of the Boeing 777 aircraft.
In order to simulate gas turbine performance at both design
and off-design points including all effect of ambient conditions
and pressure losses, it would be ideal to model each major part
of the gas turbine precisely in the GTPOM tool, i.e. compressors, combustor and turbines etc. This would however be time
consuming, and the main objective of this study is optimisation
of a combined cycle plant with an existing gas turbine and
without optimisation of the gas turbine itself. Instead, data for
the GT is provided from Rolls-Royces performance software
called eTrent, and was coded to the GTPOM tool as a lumped
component. Polynomial equations to express the engine characteristics were used and the benefit with this is that it is possible
to obtain very good accuracies within restricted boundaries;
pinlet = 127 mm H 2 O
40C Tambient 50C
0 p exhaust 400 mm H 2 O
GT exhaust temperature
GT exhaust composition
H EX
1.013
17.2
1.188
1.014
HEX
17.22
11.13
750.88
177.21
15.08
11.13
1289.2
184.59
13.45
53.67
1025.6
237.43
3.755
11.02
2910.5
237.23
13.45
53.57
1140.5
261.17
3.755
11.01
2962.1
260.15
OB JECT.
251.07
59.96
17.22
11.47
439.18
104.58
34.41
1.373
435.96
103.99
17.22
1.188
3.771
11.03
2780.8
13.45
184.19
11.13
0.0155
1.188
438.08
104.5
H EX
H EX
1.02
13.45
53.81
mas s[kg/s]
p[bar]
kW
Net efficiency
%
/kWh
%
2780.8
184.19
26.9
11.01
1.022
34.41
11.01
NG
3000
13
31.46
15
1.023
ST
( NG)
GT
321
140.4
40.35
53.57
1726
268.29
222.5
100
H EX
0
3
7.51
11.01
1.023
2962.1
260.15
26.9
49.93
mass flux =
40.35
53.47
1.031
13.45
53.17
NG
3219.2
411.48
1175.8
268.15
13.45
53.47
Equation 2
This change increased the Reynolds number and thereby the
heat transfer coefficient on water side.
2.
The second adjustment that was done was that the heat
transfer coefficient seemed to be slightly too low when the configuration consists of more then one tube row. Therefore the
assumption was made that the equation1 for calculating the heat
transfer coefficient was based on one tube row, Equation 3.
According to [vii] the heat transfer coefficient is increasing
when the number of tube rows is increasing since there then is a
higher probability that the gas side exchange heat with the wter
side.
= 0,3 Re 0, 625 Pr 0,333 0,375 C
Equation 3
Table 1: Correction factor for number of tube rows.
Number of rows 1
2906.9
235.62
3000
3219.2
409.2
34.41
0.199
0
49.93
2791.2
268.17
26.9
53.17
3219.2
411.48
2335.2
59.95
3219.2
409.2
3219.2
409.2
2476
1.012
44.113
43.648
34.41
0.199
2476
1.013
250.95
59.955
The heat exchanger specific cost (Euro per unit heat transfer
area) was reviewed and has been adjusted.
15.152
15
1.043
H EX
(N G)
10
3.
ST
26.9
49.93
H EX
46
ST
HEX
-784.3
Equation 1
2891.5
228.92
H EX
15.152
15
h[k J/k g]
t[ C]
Power output
HRSG efficiency
3.755
11.03
mass flux =
CO E
1.021
H EX
146.8
1.013
755.78
177.82
In the present GTPOM the assumption that all heat exchangers consists of U-tubes where the flow is divided into
each tube.
2780.8
152.78
750.88
177.21
0.0155 2780.8
11.03 184.19
ST
1.016
781.73
184.21
1.
251.07
59.96
FU N CT.
15.08
11.03
The model is a 2 on 1 configuration, however to avoid unnecessary equations only one GT and HRSG has been modelled
and the flow has then been doubled before entering the steam
turbine.
In order to use the data from GTPRO as reference for the
Life Cycle Analysis some of the component models had to be
modified. The heat exchanger models where modified in three
different aspects:
1.033
Also some adjustments for the original GTPOM steam turbine unit model were carried out. The investment cost of the ST
however was changed and the one that it was decided to use is
based on a simple exponential cost curve decaying as unit size
grows. This equation has been found to have high accuracy
when bidding projects.
x2
Specific Cost = x1 +
x4
+
(
x
3 Y)
Equation 4
The CCGT plant in this project was modelled with an aircooled condenser that was not originally in the GTPOM library.
A simple air-cooled condenser had been created during the previous GTPOM project, but it was never implemented into
GTPOM library. This simplified air-cooled condenser model
has been utilised with modification in this project. Equations
were added to calculate the heat transfer coefficient on air-side
while the heat transfer on steam/water side was set to a fixed
value. When both the heat transfer coefficients on each side had
been calculated the overall heat transfer coefficient could be
found, and with the logarithmic Mean Temperature Difference
and heat duty the heat transfer area could be calculated. Same
type of exponential function (Equation 4) that was used for the
ST cost calculation was then used to calculate the investment
cost for the condenser based on the heat transfer area.
Another component that was used was a gas fuel compressor. The required fuel gas pressure from the gas fuel skid depends upon fuel composition, fuel temperature, ambient temperature and engine loading. Fuel temperature limits are specified to avoid any condensation in the fuel system. The fuel
should be delivered between 38C and 149C and at least 20C
above the maximum dewpoint of the fuel. The required fuel
pressure for the Trent is from [viii] simplified to an equation
based on load. This curve, Figure 2, has a good accuracy between 30% and 100% load, which is in this case sufficient because of the low power limitation for the GT model (down to
40% load).
24
15
Not
6.5
months
years
considered
%
Operating
hours/year
4077
1748
1748
749
Power output
%
100
100
70
70
Electricity selling
price cents/kWh
4.3
6
7
10
Pressure [bar]
40
30
20
10
0
30%
40%
50%
60%
70%
80%
90%
100%
Load
The O&M costs for the genset and the ST including HRSG
and condenser was added up to one cost divided by the power
output of the plant.
When the economic scenario was set up it was chosen to
optimise the plant according to the IRR. The IRR is defined as
the discount rate which sets the net present value of a series of
cash flows over the planning horizon equal to zero and is used
as a profit measure. The IRR gives the return of an investment
when the capital is in use as if the investment consists of a single outlay at the beginning and generates a stream of net benefits afterwards.
The total capital investment was calculated and compared
with the GTPRO data and the accuracy (5% lower for GTPOM,
see Table 5) was very good considering that in reality these
figures can differ by up to 20% from different tenders when
bidding project. The difference in IRR is mostly because of the
fuel cost in GTPRO was lower than that which was set in this
project.
TQ diagram
500
450
400
300
C
GTPOM
94.95%
10.3%
350
250
200
150
100
50
0
0
The optimisation of the whole power plant was done for two
different cases. Case one (DV 11) used 11 decision variables to
optimise the plant, and case two (DV 12) used one more decision variable when the plant was optimised.
RESULTS
Comparison of plant efficiency before and after optimisation, Table 6, shows that the IRR was improved by 0.4-0.5%
units by the optimisation.
Before
After,
DV 11
After,
DV 12
99.59%
99.51%
HRSG
efficiency
base
-0.21%points
-0.25%
points
71.43%
68.83%
69.15%
COE
base
-0.01
c/kWh
-0.01
c/kW/h
96.76%
10.8%
95.96%
Figure 3 shows how the investment costs for all the components have changed after the optimisation. The investment cost
for the heat exchangers was reduced by 18% compared to the
non-optimised plant. Since the gas turbine has not been optimised in this work the investment cost for the genset was set to
a fixed value and has not been changed during the optimisation
.
Investment cost breakdown
Investment cost
Construct cost
Plant engineering
Power TD
Buildings
ST
Genset
HRSG
Base case
DV 11
DV 12
Figure 3
The lower investment cost will decrease the efficiency of
the plant, however the IRR will be improved through the decrease in cost. The TQ diagram below shows the difference
before and after optimisation with 11 decision variables. The
thick lines represent the optimised case. The recovered heat in
the HRSG is less for the optimised 48 MW and 50MW heat has
been recovered in each case.
40000
50000
Capital
IRR Investment
100%
10.3% (base)
10.7%
30000
Base
DV 11
DV 12
m2
Net
efficiency
20000
kW
10000
LTE
IPE2
IPB
Figure 5
CONCLUSIONS
The result from the optimisation shows that both the HP
pressure and IP pressure were increased from 53 (HP) and 11
bar (IP) to 83 (HP) and 13 bar (IP) respectively. The LP pressure decreases from 1.19 bar and goes towards the lowest value
1 bar, which is close to ambient pressure. This is what we expected since it needs, less steam from the boiler to the deaerator
(lower saturation temperature). To receive a high HRSG efficiency the temperature difference should be kept low and the
area of the heat exchangers large, however a large area also
means a large investment cost and it is therefore a balance between cost and efficiency that needs to be considered. The bottoming cycle optimisation resulted in a total capital cost reduction of the bottoming cycle of up to 3% (HRSG 18% and steam
turbine 1.5%).
The improvement of the IRR through the optimisation in
this present work has remained much more modest degree,
0.4%-points to 0.5%-points, compared to the previously presented IRR improvements by the case studies in the GTPOM
programme [ix]. There are several reasons for this:
The gas turbine capital cost has been dominating the majority of the total capital investment cost of the whole
plant. The previous GTPOM programme optimised the
topping cycle as well but this has not been done in the optimisation in this present work because the gas turbine
component was agreed to be modelled as a lumped component model in order to calibrate the GTPOM tools component models for bottoming cycle with reference CCGT
plant information.
Although the Trent 50 model has been successfully implemented and the component models for the bottoming cycle part
in the GTPOM tool library have been calibrated and improved
through this present work, the above results and findings has
been conversely implying that optimising whole CCGT plant
including optimisation of gas turbine itself is important in order
to drastically improve IRR of a CCGT whole plant.
ACKNOWLEDGMENT
I would like to thank those who have helped me here at
Rolls Royce and also the people at Lund University that have
been to great help during this project.
vi Mesbahi E., et. Al, An online and Remote Sensor Validation and Condition Monitoring System for Power Plants, Proceedings of CIMAC Congress 2001, 99. 833-842, Hamburg,
Germany