Applied Thermal Engineering 91 (2015) 496e506

Contents lists available at ScienceDirect

Applied Thermal Engineering

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

Research paper

Development and validation of a dynamic simulation model for a large

coal-fired power plant
Ralf Starkloff a, *, Falah Alobaid a, Karl Karner a, Bernd Epple a, Martin Schmitz b,
Felix Boehm b
a
TU Darmstadt, Institut für Energiesysteme und Energietechnik, Otto-Berndt-Str. 2, 64287 Darmstadt, Germany
b
Alstom Power GmbH, Boveristr. 22, 68309 Mannheim, Germany

h i g h l i g h t s

Detailed modelling of large coal fired power plant with minimal boundary conditions.

Validated against operational data showing a very good agreement.

Maximum relative errors of less than 5% regarding the mass flow rate, temperature and pressure.

The model will be ideal for research regarding innovative flexibility measures.

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

Article history: The rapid expansion of renewables due massive use of subsidies in the last years changes the operation
Received 21 April 2015 of fossil fired power plants in Germany dramatically. Since the large-scale storage of electricity is still
Accepted 11 August 2015 difficult and forecasts for wind and solar energy have unavoidable uncertainties, the conventional power
Available online 20 August 2015
plants have to compensate the demand in the grid. Several new measures are introduced by power
plants operators in order to respond to these new market conditions like the reduction of the low load
Keywords:
operation of the nominal load and the continuous operation at an off-design operation point. In this
Coal-fired power plant
work, a full-scale dynamic model of large scale coal-fired power plant has been developed to investigate
APROS
Dynamic simulation
the operation flexibility. The dynamic model, developed in the process simulation program APROS, in-
Validation cludes all power plant components and its associated control schemas. The only boundary conditions are
Flexibility the coal composition, the ambient temperature and the temperature and amount of cooling water in
condensers. The developed model is validated against operational data from the real power plant during
part load transients, showing a very good agreement. The validated model is of high relevance for further
investigations regarding flexibility increase in conventional large coal-fired power plants.
© 2015 Elsevier Ltd. All rights reserved.

1. Introduction sudden increase of more than 10 GW wind energy production

Coal-fired power plants are of high importance for the secure
electrical power generation in the next decades. However, the
massive growth of renewable energy in some countries changes the
market conditions [1]. In 2014, the installed capacity of wind and
photovoltaic energy has reached an amount electrical capacity of
40.4 GW and 38.2 GW, respectively [2]. Especially the fluctuating
wind energy production can lead to extreme situations, like a
* Corresponding author. Tel.: þ49 6151 16 76197; fax: þ49 6151 16 5685.
E-mail address: ralf.starkloff@est.tu-darmstadt.de (R. Starkloff).
within a couple of hours, resulting in negative prices for the spot
market [3]. Traditionally large coal-fired units were designed for an
operation of 6000e8000 full load hours per year. Due to the new
requirements of the market, the economic operation must be
secured at 1500e3000 full load hours per year [4]. It is also ex-
pected that the flexibility for conventional power plants will in-
crease even more in the near future. As a result, the power plant
operators are forced to develop new technical solutions.
The new challenges of coal-fired power plant operation can be
split in three parts. Firstly, faster positive and negative load tran-
sients between operational points are necessary due the volatile
power generation from renewable sources. Furthermore, the

http://dx.doi.org/10.1016/j.applthermaleng.2015.08.015
1359-4311/© 2015 Elsevier Ltd. All rights reserved.
 R. Starkloff et al. / Applied Thermal Engineering 91 (2015) 496e506 497

Nomenclature rad radiation

va valve
F force/volume in N/m3 w wall
f friction factor
h static enthalpy in kJ/kg Abbreviations
h0 stagnation enthalpy in kJ/kg APH air pre-heater
K correction factor APROS advanced process simulation software
Nu Nusselt number BC boundary condition
p pressure in MPa CP circulation pump
Pr Prandlt number CCV circulation control valve
Q_ heat flow/volume in kW/m3 CRH cold reheat
Re Reynolds number E extraction
T temperature in K ECO economizer
t time in s ESP electrostatic precipitator
u longitudinal fluid velocity in m/s EVAP evaporator
z longitudinal distance in m FDF forced draft fan
a heat transfer coefficient in W/(m2K) FWP feed water pump
G mass transfer in kg/(m3 s) FWPT feed water pump turbine
r density in kg/m3 FWT feed water tank
s StefaneBoltzmann constant in W/(m2 K4) G generator
c void fraction HPT high pressure turbine
HRH hot reheat
Subscripts IDF induced draft fan
con convection IPT intermediate pressure turbine
fl form loss LPT low pressure turbine
g gas phase PAF primary air fan
grav gravitation PH pre-heater
HE heat exchanger RH re-heater
i interface between phases SCAPH steam coil air pre-heater
k phase (either liquid or gas) SCR selective catalytic reduction
l liquid phase SG steam generator
pu pump SH super heater

dynamic of start-up processes (cold, warm and hot) has to be
increased in response to a sudden loss of energy generation from
wind turbines in the grid. Secondly a broader operational range is
necessary. The technical minimal load limit has to be re-evaluated.
Lower load limits save fuel and reduce the number of shut-offs and
start-ups. For combined heat and power plants a total shut-down is
often not an option since the heat demand is de-coupled from the
fluctuations in the grid. For coal-fired power plants the technical
low load limit is defined by the firing concept, burners, fuel, steam
generator design and the corresponding wateresteam cycle.
Thirdly a thermo-economical optimization of the plant within the
complete operational range, especially in the part load, is necessary
since the time running at full load is reduced. A coal-fired power
plant with lower efficiency losses at part loads may have a better
position in a liberated market.
With increasing pressure on the electricity price investment de-
cisions have to be chosen carefully [5]. A promising method is the use
of computer aided process simulation tools. In the past, steady state
process simulation programs, which are limited to time independent
operational points, were the preferred choice for process optimiza-
tion. For the above-mentioned fluctuating respond requirements of
coal-fired power plants a steady state approach is no longer suffi-
cient and a dynamic process simulation approach becomes neces-
sary. These dynamic models enable the researchers to test new
control concepts, to modify load transients characteristics and to
optimize the plant behavior in off-design operation conditions.
The first developed dynamic models contained only plant sec-
tions for a specific application. The danger of boiler implosions, in
case of a fuel or a fan trip, led to the development of mathematical
models for the air/gas dynamics of the power plants [6,7]. Whereas
other models investigated the behavior of steam generators [8,9],
turbines [10] or both [11,12]. Recently, a study has been carried out
to couple a process simulation model with a computational fluid
dynamic model in order to evaluate the effects of the three
dimensional firing system to the water steam cycle in generic coal
power plant [13]. Also, a large model of a conventional drum-based
coal-fired power plant was proposed in Ref. [14]. It is validated
against four steady state operational points, which leaves the
question open how well the model performs at transients at all.
Apart from this the focus of dynamic simulation of coal-fired power
plants shifted towards carbon capture and storage and especially to
oxyfuel combustion. Here dynamic simulation tools are mostly
used to evaluate the switch over from air to oxygen combustion or
load changes. Since no large scale oxyfuel fired plants exist,
measured data is not available either. Several researchers have
published work, with strong focus on the steam-generator gas-side.
Some of them used advanced process simulation software APROS
for model development [15,16], while some used other software
[17,18].
The objective of this study was the development of a full scale
dynamic model of an existing German power plant, including all
plant components and all required automation processes with
minimum boundary conditions, based on the real design specifi-
cations, construction drawings and control schemes. On the
wateresteam side, the only boundary conditions are inlet tem-
peratures and mass flow rates of the cooling water into the
498 R. Starkloff et al. / Applied Thermal Engineering 91 (2015) 496e506

condensers. On the flue gas path, the coal composition and the air
ambient temperature are only set as boundary conditions. Ac-
cording to our best knowledge, there are no studies up to date that
have been presented a detail dynamic model of a coal-fired power
plant in the literature. Measurement data for the response of the
most important parameters during part load operation is applied to
model validation. Aided by the validated model, old thresholds and
material limits can be re-evaluated and re-assessed.
condensation or boiling are represented by the terms c and G,
respectively. In the momentum equation, the last three terms Fva, Ffl
and DPpu correspond to the force per volume due to valve friction,
the force per volume due to friction from form loss and the pressure
change per length resulting from the pump. The terms Fik and Q_ ik
denote to the interfacial friction and heat transfer, respectively.
Both terms are calculated by empirical flow dependent correlations
which are described detailed in Refs. [24,25]. The friction between a
single-phase and the wall Fwk can be expressed as:

2. Mathematical model f k rk uk juk j

The used dynamic simulation tool is the software APROS, which
is developed by the Finnish research centrum VTT together with
Fortum [19]. APROS offers two thermal hydraulic modules, i.e. the
homogenous and the heterogeneous flow model. Both hydraulic
models solve non-linear differential equation system that repre-
sents the power plant. The high accuracy of the used mathematical
model in APROS is confirmed throughout the entire range of
operation [20,21]. Furthermore comparisons with other dynamic
simulation tools show that APROS can reproduce the real behavior
of the power plant with more accuracy [22,23].
The one dimensional homogenous model is based on the three
conservation equations for mass, momentum and energy:
vr vðruÞ
þ ¼0
vt vz
Fwk ¼ 2 (7)
D
where the terms D and fk refer to the diameter and the friction
pressure loss coefficient for the phase k. For the wall heat transfer
Q_ in the energy equation, there are three heat transfer zones with
ik
separate correlations for: wetted wall, dry wall and a transition
zone between wetted wall and dry wall. In the wetted wall corre-
lation, only the liquid phase touches the wall of the flow channel.
For the dry zone, only the gas phase is in contact with the wall. For
the heat transfer zone the critical heat flux and the minimum film
boiling temperature are needed.
The released chemical bond energy is transferred to the water/
steam side via two heat transfer mechanisms, the convective heat
transfer that is dominating in the upper (convective part) of the
steam generator and the radiative heat transfer that is dominating
(1)
in the furnace. The convective heat transfer Q_ in the steam
con

2

generator is calculated with the following equations:
vðruÞ v ru vp
þ þ ¼ Fgrav þ Fw (2)
vt vz vz Q_ con ¼ acon AHE DT (8)

vðrh0 Þ vðruh0 Þ vp with the convective heat transfer coefficient:

þ ¼ þ Q_ w (3)
vt vz vt
_ 0:6
acon ¼ Kcon m (9)
The terms Fgrav, Fw and Q_ w stand for the gravitational acceler-
ation force per volume, the friction force per volume and the heat The term AHE refers to the area of the heat exchanger and DT to
flow through walls, respectively. The term h0 in the energy equa- the temperature difference. Kcon is a factor which takes plant spe-
tion is the total enthalpy (stagnation enthalpy) including the kinetic cific parameters like the geometry into account. The radiative heat
energy. The homogenous model assumes that the physical values of transfer in the furnace can be described as:
steam and water are the same. Therefore, it is used only where the
present of an one-phase flow is guaranteed, e.g. economizer, gas Q_ rad ¼ arad AHE DT (10)
side or water pre-heaters. For two-phase flow regions in the power
plant (evaporator, condensers) the heterogeneous slip model of where Q_ rad refers to the transferred heat flow due radiation Tgas to
APROS is used [24,25]. Here the one dimensional conservation the temperature of the gas and Tw to the temperature of the wall.
equations for mass, momentum and energy are solved for both The radiative heat transfer coefficient is determined as:
phases:
T4gas  T4w
vðck rk Þ vðck rk uk Þ arad ¼ Krad s (11)
þ ¼ Gk (4) Tgas  Tw
vt vz
  here, s is the StefaneBoltzmann constant and the factor Krad takes
2
vðck rk uk Þ v ck rk uk vp the emission and view factors into account.
þ þ ck ¼ Gk uik þ ck Fgrav;k þ Fwk þ Fik The finite volume solution method is applied to solve the one-
vt vz vz
þ Fva þ Ffl þ DPpu dimensional partial differential equations in APROS. The partial
differential equations are discretized with respect to space and
(5) time. The non-linear terms are linearized. For the space dis-


cretization, the staggered discretization scheme has been applied.
v ck rk h0;k v ck rk uk h0;k vp Here, the state variables (such as pressure, enthalpy and density)
þ ¼ ck þ Gk h0;ik þ Q_ wk þ Q_ ik
vt vz vt are calculated in the center of the control volume while the flow
þ Fik uik related variables (such as velocity) are calculated at the border of
two connecting control volumes. The advantage of the staggered
(6)
arrangement is the strong coupling between velocity and pressure.
The subscript k refers to l ¼ liquid or g ¼ gas. The subscript ik As a results, some convergence problems and oscillations in pres-
denotes to the interface between two phases, while the subscript sure and velocity fields (checkerboard oscillations) can be avoided
wk stands for the interface between one phase and wall. The void [26]. In a control volume, the quantities are averaged over the
fraction and the mass exchange rate between the phases due whole control volume. The implicit method is applied for temporal
 R. Starkloff et al. / Applied Thermal Engineering 91 (2015) 496e506
discretization, where the pressure, void fraction and enthalpy
linear equation groups are solved one after the other. The density is
updated as a function of the solved pressure and enthalpy.
3. Coal-fired power plant
3.1. Reference power plant description
A coal-fired pulverized power plant consists of an air/flue gas
system and a water/steam system while the steam generator is the
interface between them. Within the steam generator the chemi-
cally bonded energy of the coal is converted into heat in a com-
bustion process. Afterwards the released heat is transferred to the
working fluid and used to generate steam for a Rankine cycle. Even
though the working principle of all coal-fired power plants is more
or less the same, every power plant is rather individually engi-
neered. The time of commissioning, fuel, control philosophy, size,
environmental regulations, firing and furnace design, number of
pre-heater and re-heater-stages and many other factors make each
plant unique. In order to build up a numerical model and start to
investigate the new operational conditions of a German power
plant, a suitable reference plant has to be chosen. The chosen power
plant design offers itself for two main reasons, on the one hand four
units at the plants Heilbronn, Bexbach, Bergkamen and Scholven
are based on this design and possible improvements may be
supporting
tubes/SH1
ECO
RH1
SH3
RH2
SH4
SH2
membrane
walll/EVAP
tangential
firing
furnace
pulverizer
Fig. 1. Sectional view of the steam generator.
499
adapted easily. On the other hand the low load operational limit for
this power plant design has been successfully reduced below 15% of
the nominal load [13]. The old low load limit of around 27% for pure
power generation was limited to minimal pulverizer load (around
50%) during two mill operation in the tangentially fired furnace
with four firing levels. The exact low limit is dependent on the
composition and ash amount of used coal blend. The reduction of
the low load limit was possible by introducing a stable and secure
single mill operation mode. The steam generator, of which a
sectional view is given in Fig. 1, was designed for domestic German
coals at a thermal power input of 1860 MW and the electrical
output of 750 MW. Currently imported world market coal blends
are fired in the plant. During a retrofit in 2009 a power increase in
the steam generator, the pulverizer system, the primary air system
and the steam turbines was achieved. As a consequence, the
nominal steam mass flow rate was increased over from 606 kg/s to
655 kg/s.
3.2. Numerical model description
One key aim of this study was the generation of a detailed dy-
namic numerical model of an existing power plant. Here, the dy-
namic calculation of the physics and a detailed control system
should represent the real plant while limiting the computational
time within a reasonable range.
+ 120.5 m
ECO: economizer
RH: re-heater
SH: super heater
EVAP: evaporator
APH: air pre-heater
SCR: selctive catalytic
reduction
SCR
APH
0.0 m
500
An overview of the air respectively flue gas path is illustrated in
Fig. 2. The ambient conditions such as pressure, temperature and
humidity are specified as a boundary condition (a).
The secondary air is supplied by two parallel working forced
draft fans (b), which is specified as an axial fan with controllable
blade angle. Afterwards the secondary air passes a steam coil air
pre-heater (c) and the regenerative air pre-heater (d) before
entering the furnace (e). The primary air is supplied by the speed
controlled two radial primary air fan (f) and flows through the
regenerative air pre-heater (d) into the pulverizer (g). The now coal
dust loaded air is then fed into the furnace and burned. In order to
consider the false air, which is unavoidable in a balanced-draft
system, ambient air is introduced at the hopper (k) of the steam
generator and adjusted for the nominal case. In Fig. 2, the
arrangement of the heat exchangers within the flue gas path is also
shown. After exiting the steam generator, the now 350e400  C hot
flue gas flows through the selective catalytic reduction (h), the
regenerative air pre-heater (d), the electrostatic precipitator (l) and
the induced draft fan (i), which keeps the furnace at a slight
negative pressure. The three different fan types should be imple-
mented according to their specifications, including a detailed
characteristic maps for a reliable part load performance. Between
the fans and the steam generator the duct system is modelled ac-
cording to the parameters length, cross section and elevation.
Additional pressure loss due bends, branch connections or dampers
is implemented via the corresponding form loss coefficients. The
pneumatic transport of the coal between the mill and the burner
adds another pressure loss, which is estimated for the design point
and implemented with a tuning of the form loss coefficient. Inter-
connecting ducts, dampers and controls should be kept in mind
when multiple gas tracks are modelled. The mill (g) is basically a
mixing point of the coal and the primary air. In this mixing the
water content of the coal is evaporated. For the dynamic behavior of
a coal pulverizer, a simple transfer function model is implemented
which is based on work of Lee [27]. Complex pulverizer models, e.g.
Fan [28] or Zhou [29] can also be used in the developed power plant
model. Flue gas cleaning devices like the electrostatic precipitator
or the selective catalytic reduction unit are modelled as pressure
drops with the correct mass. In order to simulate the steam coil air
ECO
supp. tubes / SH 1
RH1
SH3
membrane wall / evaporator
RH2
SH4
SH2
(k)
R. Starkloff et al. / Applied Thermal Engineering 91 (2015) 496e506
pre-heaters or regenerative air pre-heater, the use of the standard
APROS library components is sufficient, nevertheless for special
simulation topics, e.g. regenerative air pre-heater trip, the model
becomes more complex.
The flow path of water or steam within the steam generator is
illustrated in Fig. 3. Here, the feed water enters the economizer
(ECO) at the top of the steam generator. Afterwards the water is
directed to the membrane wall, where the evaporation takes place.
Due the once-through design the evaporation zone is not defined
and the two-phase region is load dependent. For low part load
operation, during start-up or shut-down of the plant a forced cir-
culation is in operation. A cyclone is installed to separate the
condensate droplets within the steam flow and a separator storage
tank is used for phase separation. The circulation is forced by a
circulation pump (CP) and the level is controlled by a circulation
control valve (CCV). Furthermore, a pump protection and a sub
cooling injection to prevent cavitation within the pump are
implemented. The circulated mass flow re-enters the steam
generator at the economizer inlet. The produced steam, saturated
or partly superheated, enters the steam cooled supporting tube
system (super heater 1) of the convective heat exchangers and
flows downwards to the super heater 2 (platen super heater). After
the platen super heater the steam flows through the super heaters 3
and 4. In between the super heaters three attemperator stages are
arranged for steam temperature control. The re-heater is divided
into two heat exchangers and has one attemperator in between.
The complete APROS model of the coal-fired power plant is ar-
ranged into different subsystems, like air-path, flue gas path, tur-
bine, low pressure pre-heaters, high pressure pre-heaters, feed
water pump with turbine and the steam generator, of which the
latter one is used as an example to illustrate the power plant in
APROS (see Fig. 4). Here, the connection points to other modelled
subsystems are marked with a circle. The points (a) and (b) are
considered as slave copies within the steam generator subsystem.
Each flue gas volume in the convective area has three parallel
corresponding heat exchangers on the water steam side, a part of
the membrane wall, a part of the supporting tubes and one
convective heat exchanger bundle. The heat exchanger components
and the flue gas segments are further spatial discretized.
S (h)
R (l)
C (i)
ESP BC
(e) (d) A
P SCAPH BC
H (a)
(c)
(b)
(f)
(g) (j) (2x)
SCR: selective catalytic reduction
: control damper
BC
APH: air pre-heater
BC : boundary condition SCAPH: steam coil air pre-heater
(4x)
SH: super heater
: fan RH: re-heater
BC ECO: economizer
: pulverizer
ESP: electrostatic precipitator
Fig. 2. Scheme of the gas-side.
 R. Starkloff et al. / Applied Thermal Engineering 91 (2015) 496e506 501

3.3. Control structure of the power plant model

to HP turbine to IP turbine
SH 4 For suitable simulation results of transient process simulation,
RH 2 special care has to be taken of the power plant control loops. There
attem. 3 are two basic approaches how to implement the control loops; the
RH attem. first possibility is e if available e the adaption of the controls from
SH 3 the real power plant. The other option is to use of published stan-
from FWP dard control concepts [30,31] and modify them to the specific
attem. 2 needs. Keeping the importance of the control logics regarding the
platen SH / RH 1 dynamic plant behavior in mind, the authors used as much input
SH 2 from the real power plant as necessary. Nevertheless, real control
attem. 1 schemas can be rather complex, since they take all possible even-
tualities during plant operation into account. Since a development
supporting from HP pre-heaters from HP turbine of a plant simulator was not the aim in the study, the control
tubes / SH 1 schemas were adapted to the specific needs for the simulation
scenarios. For the gas side, a forced draft fan control was imple-
separator
mented, which regulates the gas pressure downstream of the APH
membrane wall/ at a load dependent set point by adjusting the blade angles. Similar
evaporator separator storage tank
to the forced draft fan, the primary air radial fans control the gas
drain sub-cooling pressure in the primary air duct downstream of the APH by
manipulation the fans speed. In a typical tangentially fired steam
ECO CP min. flow valve generator with four firing levels, the secondary air mass flow
through each burner is, generally, controlled separately. In this
from HP
pre-heaters CCV work, the secondary air mass flow is controlled for each level due
CCV: circulation control valve the one-dimensional approach. The set point for the total second-
: valve CP: circulation pump ary air mass flow in a level is defined by the feeder speed of the
HP: high pressure corresponding pulverizer and is corrected with the oxygen content
: heat exchanger
IP: intermediate pressure in the flue gas. The primary air mass flow to each pulverizer is
: pump ECO: economizer defined by the corresponding feeder speed. On the flue gas side, the
SH: super heater induced draft fan control is active. The main task of this control
: attemperator RH: re-heater
schema is to regulate the furnace pressure by controlling the blade
Fig. 3. Overview of the heat exchangers in the steam generator and the circulation angles. Several level control circuits are modelled on the water
system. steam side, for examples the two condensers, the six pre-heaters
and the feed water tank. The steam temperature in the SH and
RH section is controlled by water injection applying different
attemperators. The feed water control circuit plays a key role for
The main process scheme of the wateresteam cycle is illustrated dynamic simulations. During once-through operation, the set-point
in Fig. 5. The steam leaves the once-through steam generator and of feed water mass flow is load dependent and corrected with
flows through the high pressure turbine (HPT) before re-entering enthalpy at the evaporator outlet and HP attemperator injection
the steam generator via the re-heater (RH). Afterwards the steam mass flow [30]. During forced circulation operation, the circulation
flows through one intermediate turbine (IPT) and two low pressure control circuit is active and keeps the evaporator mass flow at its
turbines (LPT1 and LPT2) into the condenser. The condensed water minimal value. The flue gas temperature at the cold outlet of the
is fed through seven feed water pre-heaters back to the steam regenerative air pre-heater is controlled via steam mass flow into
generator. For the power supply of the feed water pump (FWP), a the steam coil air pre-heater and should be kept 10e20 K above the
steam propelled turbine with a separate condenser is used. The acid dew point. Several other control systems are also implemented
modelling of the steam turbines is based on the given character- in the developed power plant model like, pressure control in the SH
istics for different load cases. For accurate calculation of the heat and RH sections, turbine bypasses, pump protections etc.
transfer in the heat exchangers like condenser, the four low pres- Since all required automation processes are included in the coal-
sure pre-heaters, feed water tank and two high pressure pre- fired power plant model, various number of transients can be
heaters, the implementation of the real geometries is necessary. investigated by simply changing the load demand set point. The
The condensate, feed water and drain pumps are modelled ac- developed numerical power plant model is capable of load cycling
cording to their given characteristics, this is important for the speed between low load and full load, with one, two, three or four pul-
controlled feed water pumps. The thickness, flow area and the verizers in operation.
length interconnection piping between the components should be
implemented correctly for an accurate calculation of the thermal 4. Simulation results and model validation
inertia and flow speed.
The result of the modelling procedure is a complete and detail To evaluate the model and its implemented controls, different
model of a large hard coal-fired power plant. Through careful steady state simulations were carried out before starting with
development, it was possible to minimize the number of boundary complex dynamic simulation. For this purpose, a reference design
conditions. The only implemented boundary conditions on the point (100%) was selected and the APROS model was tuned to this
wateresteam side are the inlet temperatures and mass flows of the specific point. After the adjustment to this point, the parameters
cooling water in condensers. For flue gas path, the coal composition have been kept constant for all following calculations. In a second
and the air ambient temperature are also set as boundary condi- step, another load point (105%) was calculated. The results show
tions. Even though these boundary conditions are assumed to be that the maximal relative deviations between the manufacturer
constant, they may also be time dependent. design data and APROS regarding the enthalpy and the pressure in
502 R. Starkloff et al. / Applied Thermal Engineering 91 (2015) 496e506

cyclone

= connection to other systems

exit to SCR

ECO (a)
sg inlet
(b) cold RH
CP
RH1
hot RH attemp. RH

SH3
attemp. 3

(b)
RH2

to HP turbine
SH4
attemp. 2
gas side
(b)
SH2

attemp. 1
(a)
(b)
over fire air
supporting tube system / SH1

level 4

level 3

level 2

membrane wall
level 1

Fig. 4. The steam generator system in APROS.

the steam generator were approximately 0.53% and 0.17%,
respectively.
For the evaluation of the gas path, operational data at different
load points, 60%, 80% and 100% have been used. In Fig. 6, the
pressure difference between secondary air path and ambience are
plotted. The pressure losses in the duct system are visible and show
not only a qualitative compliance with the measurement, but also a
quantitative match at certain positions. Unfortunately, no pressure
measurements are installed in the real power plant at some posi-
tions. During the steady state simulations the control system was
active, which is made visible by a negative pressure in the upper
part of the furnace. The pressure loss between the furnace and the
outlet of the air pre-heater (APH) is slightly underestimated, which
might be explained with unknown fouling amount in the
convective heat exchangers and the flue gas duct system. The
model quality of the gas side and the corresponding control logics
will be important in future investigations of fan of pulverizer
malfunction during extreme low load and one firing level in oper-
ation. Here, little experience is available and the dynamic model
may help to evaluate those scenarios.
Steady state simulations can describe the model quality
regarding heat, mass balances and correct dimensioning of the
ducts, pipes and heat exchangers. For a satisfying numerical full
scale model of a coal-firing power plant, which should be used to
investigate flexibility improvements, a dynamic comparison with
experimental data is unavoidable. Therefore, an appropriate load
change scenario has to be identified. The requirements of such a
scenario are: a large load change, change from once-through to
 R. Starkloff et al. / Applied Thermal Engineering 91 (2015) 496e506 503

SH RH

HPT IPT LPT 1 LPT 2 G

SG

E7

E4
E6
E3 cond-
E5 enser
BC
FWT=PH5 E2
PH7
BC
FWPT E1
FWP FWP BC
cond.
PH6 (e) (t) BC
cond.
pump

PH4 PH3 PH2 PH1

SG: steam generator FWT: feedwater tank : pump

SH: super heater PH: pre-heater
: heat exchanger
RH: re-heater G: generator
FWP: feedwater pump HPT: high pressure turbine BC : boundary condition
(t): turbine driven IPT: intermediate pressure turbine : valve
(e): electrical driven LPT: low pressure turbine : turbine
E1-7: extraction 1-7 FWPT: feed water pump turbine

Fig. 5. Scheme of the water/steam side.

forced circulation, start or stop of a firing level. With these the plant at the desired thermal power input. Together with the
boundary conditions and the available data of the plant operator a next load set point reduction from 42% to 33.4% after 162 min firing
load change from 100% to 22.7% has been selected. The scenario is level 2 is shut down. Ten minutes later the forced circulation starts
shown in Fig. 7. The here performed load change is unusual due its and in the following the load set point is reduced two times more
absolute length and due the fact that it is more a sum of different from 33.4% to 29.7% and then to 22.5% at minutes 175 and 185,
load changes. However, the operator has different criteria to run the respectively. For the dynamic simulation, the described events of
plant, like electricity price, sold minute reserve and heat demand. the load change scenario (see Fig. 7) have been used as an input.
Nevertheless, the scenario offers several ramps and low part load The pressure in the HP system of the steam generator is dis-
operation to check the quality of the model. It starts with a change played in Fig. 8. Here, the pressure transients in the separator bottle
of the load set point from 100% to 67.5% after 5 min. When the set and at the steam generator exit are plotted. The difference between
point is changed again form 67.5% to 42% after 110 min the lowest the two plots represents the pressure loss in super heater system.
firing level is shut down. During a pulverizer shut down the The simulated results show a good match of the experimental data
selected feeder is running down to 0% speed, while the remaining qualitatively and quantitatively. After the first and second changes
feeders have to compensate the missing coal flow in order to keep in the load set point from 100% to 67.5% and then to 42%, the plant
moves to stationary operational points. Due the inhomogeneous
coal composition of the feed in the real plants changes in the
exp. 100 %
pressure difference to ambience in mbar

model 100 %
exp. 80 %
40 model 80 % shut off firing level 1
exp. 60 %
20 model 60 % 100
load set point / nom. load in %

shut off firing level 2

0 80

-20 60 start of circula on

-40 40

-60 20
FG APH out
FG APH in
ambience

FDF out

air APH in

air APH out

FDF in

furnace

0
0 50 100 150 200
time in min
load set point
Fig. 6. Pressure difference in the secondary air path with the ambience for three
different load cases. Fig. 7. Load change scenario from 100% to 27.5%.
504 R. Starkloff et al. / Applied Thermal Engineering 91 (2015) 496e506
pressure / nom. steam pressure in %

mass flow/ nom. steam mass flow in %

120 100

100
80
80
60
60
40
40

20 20

0 0
0 50 100 150 200 0 50 100 150 200
time in min time in min
exp. pressure after steam generator model pressure after steam generator
exp. pressure in separator bottle model pressure after separator bottle exp. steam mass flow exp. mass flow ECO inlet exp. circ. mass flow
model steam mass flow model mass flow ECO inlet model circ. mass flow

Fig. 8. Comparison of the pressure in the HP section of the steam generator.

Fig. 10. Mass flows in the steam generator.

process are unavoidable even during constant set point periods

model can reproduce the superheated steam temperature with
(between 60 min and 110 min as well as between 130 min and
high accuracy. During the shut off of the second firing level and
170 min). One the contrary, the simulation produces constant
start-up of the recirculation system, the simulated temperature is
values until the next set point change occurs. Also visible is the
for short periods too high due to the slow reacting of the related
change from sliding pressure operation to fixed pressure operation,
control circuits. The simulated water temperature at the steam
which happens during the third change in the load set point from
generator inlet agrees very well with the experimental data and
42% to 33.4%. Here, the valve upstream of the high pressure turbine
gives an impression of model quality regarding the turbine stages
starts to control the pressure in the high pressure part of the steam
and pre-heater section. This is due to the fact that the turbine
generator at its minimal value. A little bit earlier, i.e. after 120 min
extraction mass flow rates towards the pre-heaters are a result of
and around 55% of the nominal hot re-heat pressure (illustrated in
the pressure difference between turbine and the pre-heater
Fig. 9), the control valve before the IP turbine controls the pressure
condenser.
in the re-heater at the minimal value. Similar to the pressure in the
After discussing the water/steam side, the view is now shifted to
HP system, the simulated reheated pressure shows quantitative
the flue gas side of the power plant. As mentioned before the flue
match against the measurement, suggesting that the implemented
gas system consists of two parallel working tracks, i.e. two selective
feed water pump and its corresponding control as well as the fuel
catalytic reduction units, two air pre-heaters, two filters and two
control and its dynamics are working correctly.
induced fans. Unlike the numerical model, where these two paths
In Fig. 10, the simulated and measured mass flow rates in the
are completely identical, differences are to be expected in the real
high pressure circuit of the steam generator are in a good corre-
unit. Therefore, Fig. 12 contains the measured flue gas pressure
spondence. Three mass flow rates, including the steam to turbine,
after the two air pre-heaters and the simulated flue gas pressure
the water at the economizer inlet and the recirculated water reveal
after one air pre-heater. As already discussed before, the absolute
in detail the behavior of the minimal evaporator mass flow control
gas pressure is hard to predict, since fouling or air leakage are
and the recirculation control. When the circulation pump is started,
changing during the operation. A qualitative match during the part
the feed water control and the recirculation control work together
load operation can be observed. It should be mentioned here that
to keep the minimal evaporator flow and the level in the bottle at
the good prediction of the pressure gradient underlines the correct
their limits. Nevertheless, there is room for improvements. The
control of the air mass flow, the air-to-fuel ratio and the accurate
start-up of the circulation system is too abrupt and should be
furnace pressure control. The air-to-fuel ratio is load dependent
smoothened, which would be left for a further stage.
and is increasing with decreasing load in order to keep the gas
In Fig. 11, the inlet and outlet temperatures of the steam
velocities high enough to enable an effective heat transfer for the
generator are illustrated. The temperature at the steam generator
convective heat exchangers.
outlet is controlled via the complex three attemperator stages. The
temperature / nom. steam temperature in %

100
pressure / nom. HRH pressure in %

120

100 80

80 60

60
40
40
20
20
0
0 0 50 100 150 200
0 50 100 150 200 time in min
time in min

exp. pressure in hot reheat model pressure in hot reheat exp. steam temperature model steam temperature
exp. pressure in cold reheat model pressure in cold reheat exp.temperature ECO inlet model temperature ECO inlet

Fig. 9. Steam pressure in the reheat system. Fig. 11. Temperatures at steam generator outlet and inlet.

pressure difference to ambience in mbar
0 evaluations together with comparisons of the simulation results
-10
-20
-30
-40
0 50 100
time in min
exp. pressure in furnace
exp. pressure at APH 2 outlet
exp. pressure at APH 1 outlet
Fig. 12. Pressure difference between flue gas and ambience.
The oxygen content in the flue gas indicates the air-to-fuel ratio
and is illustrated in Fig. 13. Due to a slightly different leakage in the
two air pre-heaters, the oxygen content at the outlets of APH 1 and
APH 2 differ considerably. The trend of simulated oxygen content at
the air pre-heater outlet follows the measurement qualitatively. It
has to be kept in mind that the overall oxygen content in the flue
gas of the plant is in between the two measurements, depending on
the exact mass flow rates. Therefore, the good agreement of the
absolute values of APH 1 outlet and the simulation should not be
overrated.
5. Conclusions
In this work, a numerical pulverized coal-firing power plant
model of an existing unit is build employing the advanced pro-
cessing simulation software APROS. The developed model includes
all mechanical components of the real plant such as, heat ex-
changers, pre-heaters, pumps, turbines, valves, condensers, fans,
dampers, air-pre-heaters, piping, tanks and the corresponding
control structures. The boundary conditions of the numerical
model have been limited to the inlet temperature and mass flow
rate of condensers, the coal composition and the ambient air con-
ditions. The developed power plant model is applied to test the
capability of APROS to reproduce the dynamic behavior of a coal-
fired power plant. Furthermore, the model is able to investigate
the low load behavior of the real plant by the reason of new re-
quirements of the conventional coal-fired power plant regarding
flexibility. Dynamic comparison studies between the simulation
results and the real plant measurements are presented for a load
20
O2 content in vol.-%
15
10
5 combustion circulating fluidized bed power plant conceptdFiring mode
0 conceptual 600 MWe oxy-combustion pulverized-coal-fired boiler, Fuel 137
0 50 100
time in min
exp. O2 content at APH 1 outlet
exp. O2 content at APH 2 outlet
Fig. 13. Oxygen content downstream of the APH.
R. Starkloff et al. / Applied Thermal Engineering 91 (2015) 496e506
150
model pressure in furnace
model pressure at APH outlets
150
model O2 content at APH outlets
505
change from 100% to 27.5% and 27.5% to 100%. According to the
with the measured data, it can be conducted that the developed
model can simulate the dynamic behavior of the real power plant
with high accuracy, even at very low loads. Minor deviations be-
tween the simulations and measurements are, however, observed.
Uncertainties, like the exact composition of the burned coal during
the experiments and the fouling level within the furnace and the
flue gas system have been a major problem for the simulation.
Another challenge for the numerical model is the estimation of the
200 difference between parallel paths of the gas side, where the
different behaviors have to be evaluated in the future studies. Due
to the high degree of detail and the implemented automation, the
developed power plant model can be used as platform for further
flexibility investigations, like faster and optimized load changes,
new control concepts during low load, primary and secondary load
support, malfunction, heat storage system, indirect firing systems
or an economizer bypass. Future studies may help other coal-fired
plants to compete in the changing market and evaluate flexibility
strategies.
Acknowledgements
The authors gratefully acknowledge the friendly support, the
supply of design data, operational measurements and technical
advice of Alstom Power GmbH and EnBW Energie Baden-
Wuerttemberg AG during the complete study.
References
[1] C. Ziems, H. Weber, Integration of large-scale renewable energy sources:
challenges to thermal generation in Germany, in: World Congress, 2011.
[2] BMWi, Zeitreihen zur Entwicklung der erneuerbaren Energien in Deutschland,
2015. Available from: http://www.erneuerbare-energien.de.
[3] M. Nicolosi, Wind power integration and power system flexibilityeAn
empirical analysis of extreme events in Germany under the new negative
price regime, Energy Policy 38 (11) (2010) 7257e7268.
[4] K. Joop, Rettungsschirm für die Energiewende, BWK J. Energy Market 64 (7/8)
(2012).
[5] J.C. Ketterer, The impact of wind power generation on the electricity price in
Germany, Energy Econ. 44 (2014) 270e280.
[6] P. Clelland, H. Kwatny, Modeling and simulation of gas dynamics in a fossil-
fuel power boiler, Instrum. Soc. Am. Adv. Instrum. 28 (1973).
[7] S.Z. Wu, et al., Dynamic modeling and simulation of gaseous systems, in:
American Control Conference, 1985.
[8] E.J. Adam, J.L. Marchetti, Dynamic simulation of large boilers with natural
recirculation, Comput. Chem. Eng. 23 (8) (1999) 1031e1040.
€m, K. Eklund, A simplified non-linear model of a drum boiler-turbine
[9] K.J. Astro
unit, Int. J. Control 16 (1) (1972) 145e169.
[10] A. Chaibakhsh, A. Ghaffari, Steam turbine model, Simul. Model. Pract. Theory
16 (9) (2008) 1145e1162.
[11] C. Maffezzoni, Boiler-turbine dynamics in power-plant control, Control Eng.
Pract. 5 (3) (1997) 301e312.
[12] H. Zindler, et al., Dynamic simulation of a 800 MWel hard coal one-through
supercritical power plant to fulfill the Great Britain Grid Code, in: Pro-
ceedings of the 6th IASME/WSEAS International Conference on Heat Transfer,
Thermal Engineering and Environment (HTE'08), Rhodes, Greece, 2008.
[13] C. Schuhbauer, et al., Coupled simulation of a tangentially hard coal fired
700 C boiler, Fuel 122 (2014) 149e163.
[14] E. Oko, M. Wang, Dynamic modelling, validation and analysis of coal-fired
subcritical power plant, Fuel 135 (2014) 292e300.
[15] R. Postler, et al., Dynamic process simulation model of an oxyfuel 250 MWel
demonstration power plant, in: 2nd Oxyfuel Combustion Conference, 2011.
[16] J. Lappalainen, et al., Modelling and dynamic simulation of a supercritical, oxy
switching case, Int. J. Greenh. Gas Control 28 (2014) 11e24.
[17] B. Jin, H. Zhao, C. Zheng, Dynamic simulation for mode switching strategy in a
200 (2014) 135e144.
[18] B. Jin, H. Zhao, C. Zheng, Dynamic modeling and control for pulverized-coal-
fired oxy-combustion boiler island, Int. J. Greenh. Gas Control 30 (2014)
97e117.
[19] VTT. APROS Simulation Software. Available from: http://www.apros.fi/.
[20] F. Alobaid, et al., Modeling and investigation start-up procedures of a com-
bined cycle power plant, Appl. Energy 85 (12) (2008) 1173e1189.
506 R. Starkloff et al. / Applied Thermal Engineering 91 (2015) 496e506
[21] F. Alobaid, et al., Dynamic simulation of a supercritical once-through heat
recovery steam generator during load changes and start-up procedures, Appl.
Energy 86 (7e8) (2009) 1274e1282.
[22] F. Alobaid, et al., A comparative study of different dynamic process simulation
codes for combined cycle power plants e part A: part loads and off-design
operation, Fuel 153 (2015) 692e706.
[23] F. Alobaid, et al., A comparative study of different dynamic process simulation
codes for combined cycle power plants e part B: start-up procedure, Fuel 153
(2015) 707e716.
[24] M. H€anninen, J. Ylijoki, The One-dimensional Separate Two-phase Flow Model
of APROS, VTT, 2008.
[25] M. Ha €nninen, Phenomenological Extensions to APROS Six-equation Model,
Lappeenranta University of Technology, Espoo, 2009.
[26] S. Patankar, Numerical Heat Transfer and Fluid Flow, CRC Press, 1980.
[27] M.S. Lee, Mathematical Model and Control of a Coal Pulverizer, University of
Austin, 1986.
[28] G.Q. Fan, N.W. Rees, Modelling of vertical spindle mills in coal fired power
plants, in: EEC 94: Electrical Engineering Congress 1994; Preprints; Enabling
Technologies, Developing Industry; Congress and Exhibition, Institution of
Engineers, Australia, 1994.
[29] G. Zhou, J. Si, C.W. Taft, Modeling and simulation of CE deep bowl pulverizer,
Energy Convers. IEEE Trans. 15 (3) (2000) 312e322.
[30] G. Klefenz, Automatic Control of Steam Power Plants, Wissenschaftsverlag,
1986.
[31] S. Dukelow, The Control of Boilers, Instrument Society of America, 1991.