Proceedings of the World Congress on Engineering 2010 Vol II

WCE 2010, June 30 - July 2, 2010, London, U.K.

Modelling Study Of Supercritical Power Plant And Parameter Identification

Using Genetic Algorithms
Omar Mohamed, Jihong Wang, Shen Guo, Bushra Al-Duri, Jianlin Wei

Abstract—The paper aims to study the whole process ip Intermediate pressure turbine
mathematical model for a supercritical coal-fired power ms Main steam
plant. The modeling procedure is based on thermodynamic
and engineering principles and the previously published m Measured
literatures. Model unknown parameters are identified using o Outlet
Genetic Algorithms (GAs) with 600MW supercritical power out Output of turbine
plant on-site measurement data. The identified parameters rh Reheater
are verified with different sets of measured plant data. sh Superheater
Although some assumptions are made in modeling process,
the supercritical coal-fired power plant model reported in the si Simulated
paper can be used to simulate the main features of the real ww Waterwall
plant once-through boiler operation and the simulation
results show the main variation trends of the process. Abbreviations:
ECON Economizer
Index Terms— coal-fired power generation, Genetic GA Genetic algorithm
Algorithms, mathematical modeling, supercritical boiler. HP High pressure
HX Heat exchanger
NOMENCLATURE IP Intermediate pressure
ff Fitness function. RH Reheater
ffr Fuel flow rate (Kg/s). SC Supercritical
SH Superheater
h Enthalpy per unit mass (MJ/Kg).
WW Waterwall
K Constant.
k Mass flow rate gain.
I .INTRODUCTION
MW MW power generated.
m Mass (Kg). The world is now facing the challenge of global
m& Mass flow rate (Kg/s). warming and environment protection. On the other hand,
P Pressure of a heat exchange (MPa). the demand of electricity is growing rapidly due to
Q& Heat transfer rate (MJ/s). economic growth and increases in population. With the
R Response consideration of environmental issues and sustainable
T Temperature (Cº). development in energy, renewable energy such as wind,
solar, and tidal wave should be only resources to be
t Time (sec). explored in theory. But the growth in demand is also a
τ Time constant (sec). heavy factor in energy equations so the renewable energy
U Internal energy (MJ). alone is not able to generate enough electricity to fill the
V Volume of fluid ( m 3 ). gap within a short time of period. Power generation using
fossil fuels is inevitable, especially, coal fired power
W& Work rate or power (MW).
generation is found to be an unavoidable choice due to its
x State vector.
huge capacity and flexibility in load following ([1], [2]).
y Output vector. The conventional coal fired power plants have a huge
l Intermediate variables vector environmental impact and lower energy conversion
℘ Input vector. efficiencies. Any new coal fired power plants must be
cleaner compared with traditional power plants.
ρ Density (Kg/ m ).
3
Supercritical power plants are the most suitable choice
ℵ Valve opening. with consideration of the factors in environmental
ε Efficiency. enhancement, higher energy efficiency and economic
Subscripts: growth. However, there has been an issue to be addressed
ec Economizer to adopt this technology in the UK because the dynamic
hp High pressure turbine response and performance still require study for better
hx Heat exchanger understanding supercritical plants in relation with
conventional subcritical plants, and the supercritical units
i Inlet
are considered by others to be unreliable in comparison
Omar Mohamed, Jihong Wang*, Shen Guo, Jianlin Wei are with the with subcritical units [3]. Therefore, the characteristics of
School of Electronic, Electrical and Computer Engineering, the
University of Birmingham, Edgbaston, Birmingham B15 2TT, UK
supercritical plants remain to be considered and
Bushra Al-Duri is with the School of Chemical Engineering, University investigated. Supercritical boilers have to be once-through
of Birmingham, Edgbaston, Birmingham B15 2TT, UK type boilers because there is not distinction between water
*
The author for correspondence. and steam phases in supercritical process so there is no
Email: j.h.wang@bham.ac.uk
need for drum to separate water steam mixture. Due to the

ISBN: 978-988-18210-7-2 WCE 2010

ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)
Proceedings of the World Congress on Engineering 2010 Vol II
WCE 2010, June 30 - July 2, 2010, London, U.K.

absence of the drum, the once-through boilers have less the reduced thermal energy steam exhausted from the high
stored energy and faster response than the drum boiler pressure turbine. The inlet temperature of the reheater is
plants. However, there are several advantages of 309 Cº and the outlet temperature is nearly 571 Cº and
supercritical power plants over traditional subcritical average pressure is 4.16MPa. The reheated steam is used
plants include: to energize the intermediate pressure turbine. Finally, the
• Reduced fuel cost due to improved plant efficiency. mechanical power is generated through multi-stage
• Significant improvement of environment by reduction turbines to provide an adequate expansion of the steam
in CO2 emissions. through the turbine and subsequently high thermal
• Plant costs less than subcritical technology costs and efficiency of the plant.
other coal technologies.
• Can be fully integrated with appropriate CO2 capture
technology.
• Fast response due to load changes and high capacity
which make them suitable for base load operation, and
also for fast load demand following.

From the literature survey, it has been found that

several models have been reported with emphasis on
different aspects of the boiler characteristics. Studying the
response and control performance of once through
supercritical (SC) units began on 1958 when work was
started on a simulation of the Eddystone I unit of
Philadelphia Electric Company and the work was extended
for simulation of Bull run SC generation unit ([4], [5]).
Yutaka Suzuki et al. modelled a once through SC boiler in
order to improve the control system of an existing plant.
The model was based on nonlinear partial differential
equations, and the simulation results indicated that the Fig.1. schematic view of the plant
model is valid ([6]). Wataro Shinohara et al. (1996)
presented a simplified state space model for SC once B. Assumptions
through boiler-turbine and designed a nonlinear controller Assumptions which are stated to simplify the process
[7]. Pressure node model description was introduced by should be logically acceptable and sufficient to transfer the
Toshio Inoue et al. for power system frequency simulation model from its complex physical state to simple
studies [8]. Intelligent techniques contributions have mathematical state. A great attention should be taken for
yielded an excellent performance for modeling. Neural the choice of assumptions because some assumptions
network has the ability to model the SC power plant with result in over-simplification and nonrelaistic system. For
sufficiently accurate results if they are trained with successful mathematical modeling, the following general
suitable data provided by operating unit [9]. However, assumptions are made:
neural network performances are unsatisfactory to • Fluid properties are uniform at any cross section, and
simulate some emergency conditions of the plant because the fluid flow in the boiler tubes is one-phase flow.
NN method depends entirely on the data not on physical • In the heat exchanger, the pipes for each heat
laws. Simulation of SC boilers may be achieved either exchanger are lumped together to form one pipe.
theoretically based on physical laws or empirically based • Only one control volume is considered in the
on experimental work. In this paper, the proposed waterwall.
mathematical model is based on thermodynamic principles • The dynamic behaviour of the air and gas pressure is
and the model parameters are identified according to a neglected.
600MW SC power plant operating currently in China. The • Only the change in internal energy is considered, the
simulation results show that the model is trustable to deviations or changes of kinetic energy and potential
simulate the whole once-through mode of operation. energy of fluid are neglected.
II. MATHEMATICAL MODEL OF THE PLANT C. Heat Exchanger model
The various heat exchangers in the boiler are modelled
A. Plant description by mass and energy balances. The sub-cooled water in the
The unit of a once-through supercritical 600MW economizer is transferred directly to a supercritical steam
power plant is selected for the modelling study. The through the waterwall without passing with the
schematic view of the boiler is shown in Fig.1. Water from evaporation status.. The equations are converted in terms
the feedwater heater is heated in the economizer before of the derivatives pressure and temperature of the heat
entering the superheating stages through the waterwall. exchanger. Refer to Appendix A for the nomenclature and
The superheater consists of three sections which are low abbreviations. The mass balance equation of the heat
temperature superheater, platen superheater, and final exchanger (control volume) is:
stage superheater. The main steam outlet temperature is
dm
about 571Cº at the steady state and a pressure is 25.5 MPa. = m& i − m& o (1)
There are 2 reheating sections in the boiler for reheating dt
For constant effective volume, Equation (1) will be:

ISBN: 978-988-18210-7-2 WCE 2010

ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)
Proceedings of the World Congress on Engineering 2010 Vol II
WCE 2010, June 30 - July 2, 2010, London, U.K.

dρ responds to a signal from PI controller. This prevents the

V = m& i − m& o temperature high fluctuation and ensures maximum
dt
The density is a differentiable function of two variables efficiency over a wide range of operation.
which can be the temperature and pressure inside the
control volume, thus we have: D. Fluid flow
∂ρ dP ∂ρ dT
V( . + . ) = m& i − m& o The fluid flow in boiler tubes for one-phase flow is:
∂P T dt ∂T P dt
m& = k. ΔP (10)
The energy balance equation: Which is the simplist from for fluid flow in boiler tubes.
dU hx
= Q& hx + m& i hi − m& o ho The flow out from the reheater and main steam
dt respectively are:
Also P
m& rh = K1′ ⋅ rh ⋅ ℵrh (11)
dU hx ⎡ ∂ρ dP ∂ρ dT Trh
= V ⎢h( . + . )+
dt ⎣ ∂P T dt ∂T P dt Pms
m& ms = K 2′ ⋅ ⋅ ℵms (12)
∂h dP ∂h dT ⎤ dP Tms
ρ( . + . )⎥ − V
∂P T dt ∂T P dt ⎦ dt The detailed derivation of (11), and (12) can be found in
Then, the energy balance equation becomes: [10].
⎡ ∂ρ dP ∂ρ dT ∂h dP ∂h dT ⎤ E. Turbine / Generator Model
V ⎢h( . + . ) + ρ( . + . )⎥
⎣ ∂P T dt ∂T P dt ∂P T dt ∂T P dt ⎦
(2) The turbine and generator can be modelled only by
dP & energy balance equation. The boiler response is very slow
−V = Qhx + m& i hi − m& o ho compared with turbine/ generator responses. Therefore,
dt
there is no need to detail the model of the
Solving (1) and (2) to get the pressure and temperature turbine/generator as in the case of SC boiler. The turbine
state derivatives, energy balance equation:
Q& + m& i H i − m& o H o
P& = hx (3) m& ms .( hms − hout ) = W& hp (13)
τ
T& = C ( m& i − m& o ) − DP& (4) m& rh .( hms − hout ) = W& ip (14)
where: The generated power of the plant:
∂h MW = ε ( W& hp + W& ip ) (15)
ρ
∂T P
H i = ( hi − h − ) (5) F. Heat transfer rate
∂ρ
The combustion heat transfer rate is a first order lag of
∂T P the fuel flow rate ([8], [11]). The heat transfer rates of the
various heat exchangers are directly proportional to the
∂h main combustion heat transfer rate. This greatly simplifies
ρ the mathematical model without sacrificing the important
∂T P
H o = ( ho − h − ) (6) dynamics to be considered for the plant.
∂ρ 1
∂T T Q& = ffr (16)
1 + τ 1s
∂ρ ∂h
ρ . Q& ec = K ecQ& (17)
∂h ∂P T ∂T
τ = V( ρ − P
− 1 ) (7)
∂P T ∂ρ Q& ww = KwwQ& (19)
∂T P
1 Q& sh = KshQ& (20)
C= (8)
∂ρ Q&rh = KrhQ&
V (21)
∂T P

∂ρ G. Final form of the model

∂P
D= T
(9) x&1 = f ( x1 ,℘, l ) (22)
∂ρ
∂T x& 2 = f ( x1 ,℘ , l ) (23)
P

y = f ( x1 , x2 ,℘,l ) (24)
The temperature of the superheater is controlled by the
attemperator. Therefore, the input mass flow rate to the where x&1 represents the pressure state vector, x&2 is the
superheater is the addition of the SC steam and the water temperature state vector, ℘ is the input vector which
spray from the attemperator. The amount of attemperator contains the direct inputs of the boiler (fuel and feedwater
water spray is provided by opening the spray valve which flows), and l is the intermediate variables vector which

ISBN: 978-988-18210-7-2 WCE 2010

ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)
Proceedings of the World Congress on Engineering 2010 Vol II
WCE 2010, June 30 - July 2, 2010, London, U.K.

are the intermediate mass flow rates of fluid. The output • Mass flow rate of reheated steam from reheater
vector y can be taken as any variable from the vectors outlet to the IP turbine.
x1 and x 2 since the model is implemented in MATLAB / • Generated power of the plant.
In recent years, Genetic Algorithms optimisation tool
SIMULINK, the observation of any variable dynamics is
has been widely used for nonlinear system identification
available at any point in the model.
and optimization ([12], [13], [14]). This is because it has
many advantages over conventional mathematical
III. MODEL PARAMETER IDENTIFICATION
optimization techniques. It has been proved that the GA
tool is a robust optimization method for parameters
A. Data selection:
identification of subcritical boiler models [11] and Tube
The data should be informative for successful
Ball mills ([15]). Initially, the GA produces a random
identification procedures. The once-through operation of
initial population. Then, it calculates the corresponding
the plant starts nearly above 30% of rated load (Fig.2.).
fitness function to recopy the best coded parameter in the
The behaviour of the plant during start-up is completely
next generation.. The GA termination criteria depend on
different from that of the once-through mode because the
the value of the fitness function. If the termination criteria
separator separate the water from steam during start-up
is not met, the GA continue to perform the three main
process and the water is re-circulated for further heat
operations which are reproduction, crossover, and
addition. After start-up process is completed, the
mutation. More details about this technique can be found
circulation valve closes to allow the once-through
in the text of David Goldberg 1989 [16]. The fitness
operation. Therefore, the model for the boiler during start-
function for the proposed model is:
up is different from the once-through mode model and the
N
start-up data shouldn't be used for identification of the ff = ∑ ( Rm − Rsi )2 (25)
proposed model parameters. Only the data of once-through n =1
operation is used for the identification and verification
which is the sum of the square of the difference between
measured and simulated responses for each of the
B. Identification procedures:
variables mentioned in this section. N is the number of
The parameters of the model which are defined by
points of the recorded measured data, The identification
from (5) to (9) and the other parameters of mass flow
scheme is shown in Fig.3. The load-up data have been
rates’ gains , heat transfer constants, turbine/generator
used for identification from 30% to 100% of load. The
parameters are all identified by genetic algorithm. Even
next step is to verify the proposed model. The model is
though some of these parameters can be found from the
verified from load down from 100% to 55% and steady
properties of water/steam or steam table and thus they are
state data to cover the large range of once-through
inherently not constant, but these parameters are fitted
operation. The GA parameters for identification of optimal
directly to the actual plant response to save time and
parameter of the model are mentioned below:
effort. Various boiler responses have been chosen for
identification and verification. First, the parameters of
Population type: double vector
pressure derivatives equations are indentified. Secondly,
Creation function: uniform
the identification is extended to include the temperature
Population size: 50
equations and turbine/generator model.
Mutation rate: 0.1
Mutation function: Gaussian
Migration direction: forward
Selection: stochastic uniform
Crossover function: scattered

Fig.2. Power survey for the modelled 600MWSC plant

Over more than 24hrs

The responses which are chosen for identification and

Fig.3. schematic diagram of model's parameter
verification are:
identification
• Reheater pressure.
• Main SC steam pressure. Some of the identification results are shown in Figs.4-
• Main SC steam temperature. 7, where the dashed line for the simulated and solid line
• Mass flow rate of SC steam from boiler main for the measured time responses. From the identification
outlet to HP turbine. results, it has been observed that the power responses are

ISBN: 978-988-18210-7-2 WCE 2010

ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)
Proceedings of the World Congress on Engineering 2010 Vol II
WCE 2010, June 30 - July 2, 2010, London, U.K.

well matched; the simulated main steam pressure and main Model response
570
Plant response
steam temperature responses are also reasonably matched.
There are so many parameters in the model so we cannot
560
mention them all. The optimal parameters of the heat

Temperature (C)
exchanger model are mentioned in Table 1 and for heat
transfer rates are mentioned in Table 2. 550

Table 1. Heat exchanges' parameters

540
HX Hi Ho C D
ECON 10.2 13.6 2.1e-6 -3.93
530
WW 12.2 13.3 -1.2e-6 -0.1299 0 100 200 300 400 500 600
SH 20.5 45.9 1e-6 -3.73 Time (min)

RH 19.8 22. -1e-6 -17.9

Fig.6 Load-up main steam temperature
Table 2. Heat transfer rate parameters Plant response
500
τ 1 (s) K ec K ww K sh K rh Model response
9.3 5.7785 7.78 23.776 21.43

main steam flow (Kg/s)

400

Plant response 300

700
Model response
600 200

500
Power (MW)

100
400

300 0
0 100 200 300 400 500
200 Time (min)

100 Fig.7 Load-up main steam flow

0
0 100 200 300 400 500 IV. MODEL VERIFICATION
Time (min)
The validation of the proposed model has been
Fig.4 Power response for plant and Model performed on the variables mentioned in section. III by a
number of data sets which are the load down and steady
30 state data. Figs.8, 9, and 10 demonstrate the reheater
Plant response
Model response pressure and power for steady state operation. Also, the
25
mainsteampressure (MPa)

main steam flow rate and power for load down operation
20
are shown in fig.6. From the results mentioned, it is
obvious that the model response and the actual plant
15 response are very well agreed for the power, reheater
pressure and steam flow rate. Thus the model is valid for
10 steady state and load change conditions and it will be very
useful for control synthesis.
5

700 Plant response

0 Model response
0 100 200 300 400 500
Time (min)
600
Power (MW)

Fig.5 Load-up main steam pressure response

500

400

300
100 200 300 400 500
Time(min)

Fig.8 Load-down power response

ISBN: 978-988-18210-7-2 WCE 2010

ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)
Proceedings of the World Congress on Engineering 2010 Vol II
WCE 2010, June 30 - July 2, 2010, London, U.K.

Apparatus and Systems, vol PAS-90, No. 4, 1971,

Plant response pp.1725-1733.
5 Model response [3] P. Kundur “A survey of utility experiences with power
plant response during partial load rejection and system
Reheater pressure (MPa)

4 disturbances” ”. IEEE Transactions on Power Apparatus

and Systems, vol. PAS-100, No.5 , June 1981, pp.2471-
3 2475
[4] J. Adams, D. R. Clark, J. R. Luis, and J. P. Spanbaur,
2 “Mathematical Modelling of Once-Through Boiler
Dynamics”, IEEE Transactions on Power Apparatus and
1 Systems, vol 84, No. 4, Feb 1965, pp. 146-156.
[5] B. Littman, and T. S. Chen, “Simulation of bull-run
0 supercritical generation unit”. IEEE Transactions on Power
0 200 400 600 800 Apparatus and Systems, vol. 85, 1966, pp.711-722.
Time(min) [6] W. Shinohara, and D. E. Kotischek, “A simplified model
based supercritical power plant controller”. Proceeding of
Fig.9 Steady state reheater pressure the 35th IEEE conference on decision and control, vol .4 ,
1995, pp. 4486-4491.
[7] Y. Suzuki and P. Sik, Y. Uchida “Simulation of once-
Plant response
through supercritical boiler”, SIMULATION, vol.33, 1979,
500 Model response
pp. 181-193.
Main SC steam flow (Kg/s)

[8] T. Inoue, H. Taniguchi, and Y. Ikeguchi “A Model of

400
Fossil Fueled Plant with Once-through Boiler for Power
System Frequency Simulation Studies” IEEE Transactions
300
on Power Systems, vol. 15, No. 4, 2000, pp. 1322-1328.
[9] K. Y. Lee, J. S. Hoe, J. A. Hoffman, H. K. Sung, and H.
200
J. Won, “Neural network based modeling of large scale
power plant”. IEEE Power Engineering Society General
100
Meeting. No (24-28), 2007, pp. 1- 8.
[10] A. S. Leyzerovich "Steam Turbines for Modern Fossil-Fuel
0
0 200 400 600 800 Power Plant". The Fairmont Press, Inc. 2007.
Time(min) [11] A. Ghaffari, A. Chaibakhsh. “A simulated model for a
once through boiler by parameter adjustment based on
Fig.10 steady state steam flow rate genetic algorithms” Simulation Model and Practice Theory,
vol.15, August 2007, pp. 1029-1051.
V. CONCLUDING REMARKS [12] J. Wei, J. Wang, and Q. H. Wu, “Development of
multisegment coal mill model using an evolutionary
A model of coal fired power generation with computation technique”. IEEE Transactions on energy
conversions, vol. 22, 2007, pp. 718-727.
supercritical boiler has been presented. The model is based
[13] J. Wang, J.D. Wang, N. Daw, Q.H. Wu, “Identification of
on thermodynamic laws and identified according to pneumatic cylinder friction parameters using genetic
operating data record. The model is also verified by algorithms”, IEEE/ASME Transactions on Mechatronics,
different data sets and the simulation results show that the vol 9. No.1, 2004, pp.100-107.
model is valid. As a future work, the model will be [14] S. Mangan and J. Wang, “Development of a novel
combined with a nonlinear mathematical model of coal sensorless road gradient estimation method based on
mill to obtain a wide mathematical model for the whole vehicle CAN Bus data”, IEEE/ASME Transactions on
process from coal grinding to electricity generation. In Mechatronics, vol. 12, No.3, 2007, pp. 375-386.
addition, a further study will be carried out to improve the [15] J. Wei, J. Wang, S. Guo, “Mathematic Modelling and
Condition Monitoring of Power Station Tube-ball Mill
plant response by control system design.
Systems”, The Proceedings of AMERICAN CONTROL
CONFERENCE, vols 1-9, 2009, pp. 4699-4704.
ACKNOWLEDGEMENT [16] D. E. Goldberg “Genetic Algorithms in Search,
Optimization, and Machine Learning” Addison Wesley.
The authors would like to thank the Libyan Cultural 1989.
Affairs for the Scholarship and also give their thanks to
EPSRC(EP/G062889) and E.ON for their research funding
support. The author also wish to express their sincere
appreciation to Professors Junfu Lv and Qirui Gao from
Tsinghua University for their contribution.

REFERENCES

[1] F. Laubli, and F. H. Fenton, “The Flexibility of The

Supercritical Boiler as A Partner in Power System Design
and Operation: Part I” IEEE Transactions on Power
Apparatus and Systems, vol PAS-90, No. 4, 1971, pp.1719-
1724 .
[2] F. Laubli, and F. H. Fenton, “The Flexibility of The
Supercritical Boiler as A Partner in Power System Design
and Operation: Part II” IEEE Transactions on Power

ISBN: 978-988-18210-7-2 WCE 2010

ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)