You are on page 1of 7

P R E P R I N T ICPWS XV

Berlin, September 811, 2008

Usage of Water & Steam Properties in Computational Intensive Dynamic Simulations

Axel Butterlin a, Dieter Schiesser b, Dr. Haiko Steuer c

Siemens Energy, Power Generation


Email: a Axel.Butterlin@siemens.com, b Dieter.Schiesser@siemens.com,
c
Haiko.Steuer@siemens.com

Transient Process Simulation has become an inevitable tool for innovative Power
Plant Development. Especially matters of unit safety in respect to material stresses of
components in the energy converting water & steam cycle process do require detailed
modeling and computational intensive dynamic simulations. In these applications the
usage of fast and reliable water & steam properties is a key factor for obtaining
meaningful results in acceptable computation time. In Siemens power plant simulation
tools special table based interpolating water & steam property functions, derived from
the IAPWS-IF97 functions, have been introduced for significant reduction of
computation time.
This work at first gives an overview of the fundamentals for transient simulation of
water & steam cycles in Power Plants and the resulting requirements for water &
steam property functions. As a typical example of computational intensive simulation
a material stress study of a Once through Benson Steam Generator is chosen where
oscillatory instability in the water & steam 2-phase region typically is studied during
Plant Engineering process. A comparative profiling of IAPWS-IF97 (Bochum) and
the interpolating table functions shows the performance increase by using pre-
processed-table functions instead of the IAPWS-IF97 formulations. Performance
increase of the IAPWS-IF97 code could be possible by implementing the availability
of partial derivates in respect to the input arguments. Another appearing requirement
is the robustness of the property functions in respect to arbitrary input arguments,
which is inevitable for numerical solving of transient flow equations.

Introduction delivers detailed information of the plant design in


an ideal steady state operation. This however, is an
Increasing prosperity of mankind up to now is operation mode which in a highly deregulated and
tightly linked to an increasing energy demand flexible power network system can not often be
especially in respect to high-value electric power. achieved. So for a better understanding of the plant
So providing clean and affordable electric Power to behavior transient simulation of the water & steam
all human beings is on of the most ambitious cycle becomes more and more inevitable.
challenges in our world. In that field technological Siemens Energy uses a set of transient
efforts lead to high sophisticated and innovative simulation tools with different detail levels in order
concepts which permit high effective and to either simulate overall plant behavior or special
environmental gentle methods of power production. effects in selected components.
Siemens Energy is providing plant solutions in a Whereas in overall plant simulation the general
wide range of power plant applications, with one transient behavior of many interacting components,
focus on conventional fired power plants. In fossil which often are highly simplified, are relevant, in a
fired power plants, as well as in nuclear plants and detailed component simulation special transient
in solar thermal systems, a sophisticated water& physical effects are studied under given boundary
steam cycle system is used as central heat conditions. In both types of simulation the transient
transmission medium. The plant engineering layout system is described by a big set of algebraic and
for this water & steam cycle is a steadily subject of partial differential equations. The normal way to
optimization according to material innovations, solve the transient simulation task is to discretize
concept improvements and thus better efficiency the space dimension in finite differences or by
and pollution control. Here in a first step steady method of lines. Although the system normally is
state calculation of the so called balance of plants reduced to one space dimension due to the time
transient a complete equations system of ordinary
differential equations (ODEs) has to be solved each Table 1: Abbreviations in the Equations system:
simulation step in time. This results in a compu-
tational effort which lies far above steady state p N/m Absolute pressure
applications. This simulation effort is combined A m Inner cross section of the tube
with a much more frequent use of water& steam
property calls which are necessary in an argument
U m inner perimeter of the tube, U=.di
representation depending on the form of the 
m kg/s mass flow rate
equation system. h J/kg specific enthalpy

Equation Systems for 1-D transient Water & qmf W/m inner (metal-fluid) heat flux
density
Steam Cycle Simulation
g m/s acceleration of gravity
The central element of the computational fluid deg inclination angle in z-dimension
dynamics is a set of flow equations, describing the L m Tube length
fluid flow in a tube (see figure 1) [1]. In a 1-D
t s time
system with its on space variable z these are the
conservation laws with its three equations repre- kg/m densitiy
senting mass, momentum and energy conservation.
Depending on the choice of the state variables the
equations can differ in their form. In this formulation other dependent thermo-
Presently in Siemens simulation tools two types dynamic properties have to be determined by water
of formulations are used, which both are types of & steam property functions by using the state
Euler equations with additional external source properties and h as arguments. Mainly these are
pressure p and fluid temperature T, but also further
terms.
material properties like specific heat capacity cp,
dynamic viscosity and heat conductivity , which
z
implicitly appear in heat transfer and friction
pressure loss correlations. However, if one neglects
di
U the term dp/dt (written in grey) only direct property
A H functions appear in the equation set and no
L
derivative of a property function.
As a consequence the Jacobian matrix, which is
necessary to solve the system dynamically and
Figure 1: This picture of the inner tube which contains all partial derivatives in respect to
cylinder explains the basic geometry the state variables does only contain first order
parameters, which occur in the derivatives of property functions.
balance equations. The term dp/dt reflects energy dissipations in
context with very fast pressure changes like
pressure shock waves and can be neglected in
nearly all normal power plant operation modes.
The first formulation is based on the states
densitiy , mass flow m
 and specific enthalpy h The situation changes in a second used
and looks as follows: formulation of the conservation laws. In this vari-
ation one takes p, m  and h as state variables and
1 m thus gets the following equation system:
=
t A z p h 1 m
+ =
m 1 m
2
p p p t h t A z
= A + g sin +
t A z z
friction
z m 1 m 2 p p
= A + g sin +
t A z
h 1 p p h friction
z z
+ = h 1 p
t h t h t =
t t
m 1 p h 1 p U qmf m 1 p h 1 p Uqmf
+ + +
A z z z friction A +
A z z z friction A

2
3500
steam

3000

sa supercritical
tur
2500 ate
d st e p = 350 bar
am
p = 100 bar
2000
h [kJ/kg]

critical point
1500 sa
tur
at e
dw
ate
1000 r water

two-phase region
500

0
0 200 400 600 800 1000
rho [kg/m^3]

Figure 2: Density-Enthalpy diagram with two isobars (a third isobar with critical pressure p = 220.64 bar is
shown in between with a dashed line). All property information of the liquid phase is concentrated
in a small band between the line of saturated water and high pressure isobars (350 bar is a state of
the art maximum pressure in power plant applications).

3000
steam

2500
rho = 200 kg/m^3
two-phase rho = 500 kg/m^3
2000 critical point
h [kJ/kg]

1500

supercritical

1000
water

500

0
0 50 100 150 200 250 300 350 400
p [bar]

Figure 3: Pressure-Enthalpy diagram with two isochors. Here the liquid phase in the relevant pressure region
is far more extended. Isochors are not concentrated to a small band

Here partial derivatives of the water & steam (water) the needed property functions have very
properties also appear in the mass balance equation high sensitivity to density changes and therefore
which cannot be neglected and therefore the often are the reason for numerical difficulties
Jacobian matrix will contain second order deri- (inevitable numerical argument errors can lead to
vatives of the property functions. big property errors). This is illustrated in Figure 2
where the liquid phase is concentrated in a small
Such, the advantage of the first formulation is band between the line of saturated water and the
that one can get rid of second order derivatives. dotted isobar of p = 350 bar.
However, in this formulation for the liquid phase

3
The second formulation uses p and h in the As a consequence the precision of the results
argument plane and thus is quite advantageous in can rarely be better than 1% in peak values and
respect to numerical argument errors (see Figure 3 about 5% in transients. However, it has to be
for comparison). remarked that the main focus of transient studies is
on the principal transient evolution of the physical
Precision aspects in transient 1-D-Powerplant properties and not on the absolute precision of the
Simulation values itself.

In order to bring a power plant system model to Detailed Component Simulation of a Fossil Fired
a level on which transient simulation results are Steam Evaporator
available in acceptable time, as much simpli-
fications as possible must be applied to the model. For illustrating the application of transient
So each transient model must be balanced between power Plant simulation a mass flow stability
calculation effort and acceptable computation time analysis of a fossil fired once-through steam
in respect to the specific issue. Following major evaporator is presented here. Figure 4 shows a
simplification are made in our 1-D-power plant typical coal fired power plant arrangement with the
simulation to describe the system: steam generator unit in the center. In this steam
generator incoming feed water is transformed to
Reduction to one dimensional system which superheated steam by firing coal inside the tube
implies an only ideal homogeneous fluid cross walled vessel. The evaporation heating surface
section in the tube. consists of a big number of parallel tubes, which are
Consideration of only one fluid component. installed helically around the combustion chamber.
This means that always water and steam have Figure 5 shows the principle of a once-through
the same velocity and temperature at the same steam generator where slightly subcooled water
location. enters the evaporator at the bottom and slightly
Friction Pressure loss and thermal heat transfer superheated steam leaves at the top of the
based on heuristic correlations determined in arrangement.
steady state test environments. In this arrangement hundreds or thousands of
Discretization of the remaining one space tubes are put in parallel and normally should have
dimension in a more or less rough grid similar mass flow rate and heat input (see Figure 6
Time integration with numerical solver and for illustration). However, in a complicated firing
thus time discretization and loss of accuracy in arrangement and engineering layout often
time dimension

Figure 4: Coal-fired power plant. Steam generator is located in the centre, on the right side turbine and
electrical equipment, on the left the flue gas section with the chimney [2].

4
- m - h formulation and by neglecting the grey
Outle t
marked property derivatives in the enthalpy
equation.
Although the system is simplified according to
the above mentioned precision aspects, a spatial
Burne rs discretization into more than 1000 nodes leads to
more than 3000 equations each requiring numerous
property functions evaluation in every time
integration step. As time integration in case of
I nl et
instability analysis often steps ahead in milli-
seconds one can imagine that a stability analysis
regarding oscillation periods of minutes can be a
challenging task for property function performance.
A typical stability analysis consists of following
Figure 5: The principle of a once-through steps:
steam generator. State initialization with steady state
simulation tool ( e.g. KRAWALmodular [4])
Dynamic relaxation of disturbances with
constant boundaries
Perturbation analysis
The typical result of a perturbation analysis is
shown in Figure 7.
The initial dynamic relaxation process is used for
IAPWS-IF97 and INTH2O benchmark analysis
below.

Inlet mass flow per tube (kg/s)


1.6

1.2

0.8
Figure 6: Typical tube arrangement of a once-
through steam generator. 0.4
dynamical disturbances in heat input can occur.
0
Depending on the size of the disturbance and on the
mechanical construction of the heating surface such -0.4
disturbances might lead to oscillatory instability of 0 500 1000
the mass flow rate. This is highly unwanted and can Time (sec)

lead to damages or even tubes ruptures due to


Outside tube wall temperature (C)
periodically alternating tube wall temperatures. 520
Such higher order instability [3] can only be
revealed by a detailed transient simulation study, 480

where the fundamentals of the thermohydraulics are at tube Length: 210 m


440
modeled for the specific arrangement. In a coal
fired once-through evaporator with a big number of 400
parallel tubes it is sufficient to model one
representative tube with a fixed inlet and outlet 360
at tube Length: 160 m
pressure. The fixed boundary pressures represent
320
the collection of parallel tubes ruling the pressure 0 500 1000
drop (nearly) independent from the mass flow rate Time (sec)
of a single tube. So the model concentrates on the
detailed representation of a single tube with exact Figure 7: Dynamic stability analysis.
geometry, heat input and especially the water &
steam flow dynamics. In this case the flow
dynamics is described with the above mentioned

5
Performance Adaptions of the Water & Steam Table 2: Comparison of calculation time for IAPWS-
Property Functions
IF97 (Bochum) and INTH2O calls.
Siemens Power Generation has adopted the IF97 INTH2O IF97/
IAPWS-IF97 formulation since 1999 and is using Function (msec)* (msec)* INTH2O
the fundamental IAPWS-IF97 source code deve- p(,h) 6110 31 197
loped by the Lehrstuhl fr Thermodynamik of the hsatwater(p) 125 16 7,8
Ruhr-Universitt of Bochum (Prof. Wagner) [5] for hsatsteam(p) 140 16 8,8
its thermodynamic design of power plants. T(p,h)/T(,h) 80 31 2,6
However, due to the mentioned requirements in Tsat(p) 16 16 1,0
relation to performance and partial derivatives in
(p,T) 63 31 2,0
detailed transient simulations not this code is used
satwater (p) 172 16 10,8
directly, but interpolating tables (INTH2O) have
been generated based on the IAPWS-IF97 property satsteam (p) 190 16 11,9
functions. (p,h) 172 32 5,4
For these tables a large set of interpolation satwater (p) 125 16 7,8
values has been stored in tables for each necessary satsteam (p) 140 16 8,8
property function together with its first order partial (p,T) 219 31 7,1
derivatives. Thus for an arbitrary argument input satsteam (p) 330 16 20,6
the corresponding property function can be Prandl-num.(p,T) 265 32 8,3
determined very rapidly by the help of the next
surrounding interpolation values. In most cases the *) calculation time for 100000 calls by scanning the
interpolation is processed linearly, which is the arguments linearly from (p: 1-211bar, T: 100-400C, h:
easiest alternative. Although here unsteady 1000-3000 kJ/kg, : 50-550 kg/m)
derivatives occur this does not lead to numerical However, the listed functions are not all used in
the same intensity. The - m
problems in most applications. Figure 8 shows the  - h flow formulation
interpolation principle with the pressure property
determination as a function of the surrounding requires a very frequent evaluation of p(,h), where
argument tupels. the difference between IAPWS-IF97 (Bochum) and
INTH2O is very significant. Furthermore partial
derivatives determination is included in the
INTH2O call whereas the IAPWS-IF97 (Bochum)
normally has no derivative information. Therefore
numerical differentiation is used for an IAPWS-
IF97 implementation, which again triples the
calculation effort for a two argument function.
As an example benchmark for an IAPWS-IF97
(Bochum) and INTH2O comparison under real
life conditions the initial dynamic relaxation step
of the above stability analysis has been chosen. In
this step the initial states which have been
determined by the help of a steady state simulation
tool and thus with another set of equations, have to
be reinitialized with the transient simulation
equation set (here the - m  - h formulation). This
Figure 8: Interpolation of variables. The cross
run, which typically starts with big initial
in the middle remarks the searched
disturbances which relaxes has to be simulated until
pressure property p(,h).
all transients disappear. The final set of states at the
end of this run is then a proper starting point for the
The following table compares the calculation
actual perturbation analysis. Simulation and
effort for the property functions which are needed
benchmark results for this relaxation run are shown
for the above stability analysis:
for both IAPWS-IF97 and INTH2O in Figure 9.

6
2.2
2.2

2.1
IAPWS-IF97 2.1
INTH2O

Mass Flow (kg/s)


Mass Flow (kg/s)
2 2

tube inlet tube outlet


1.9 1.9 tube inlet tube outlet

1.8 Time of operation : 579s Time of operation : 78s


1.8
number of time integration steps: 796 number of time integration steps: 610
1.7 1.7
0 20 40 60 80 100 0 20 40 60 80 100

Time (sec) Time (sec)

Figure 9: Simulation and benchmark results for the relaxation run of the dynamic stability analysis.

Robustness Requirements of Property Functions Literature


in Transient Flow Simulation
[1] Bird, Stewart, Lightfoot: Transport
In the above benchmark the implemented Phenomena, New York, John Wiley & Sons
IAPWS-IF97 formulation of has been used in (1960).
combination with a special extrapolation algorithm [2] EON Kraftwerke GmbH: Kohlekraftwerk
which also guarantees a result in case of input Schkopau , Infobroschre zum
arguments lying outside the permitted range. This Kohlekraftwerk Schkopau., www.eon-
algorithm has been implemented in the Siemens kraftwerke.com.
IAPWS-IF97 interface by Bennoit [6]. Also in [3] J. Franke et al.: Flow Oscillations in the
INTH2O in case of arguments out of range the Evaporator Section of Steam Generators, VGB
result is determined by extrapolation of the nearest Kraftwerkstechnik, 67, Pages 30-36, January
(1987).
valid interpolation values. This procedure is
inevitable for the numerical solving of the time [4] G. Lbel: Water and Steam Properties for
Power Plant Design, Proceedings of the 15th
evolution, because often numerical integration
ICPWS, Berlin (2008).
errors of the states lead to out of range incidents. So
[5] W. Wagner, B. Rukes: IAPWS-IF97: Die neue
in case of the above benchmark run for example
Industrie-Formulation. Brennstoff-Wrme-
following incidents during numerical integration Kraft 50 (1998), 42-47.
had to be managed:
[6] S. Bennoit: Siemens interface for IF97,
Negative Density of -86 kg/m -> IAPWS-IF97 personal communication.
returns negative pressure of -95bar
Density of 3500 kg/m -> IAPWS-IF97 returns
a pressure of 20000bar
Both reactions of this safeguarded IAPWS-IF97
can be fairly handled by the solver and do not lead
to diverging reactions. It has to be remarked that
normally these unphysical intermediate arguments
do not lead to unphysical results of the physics, but
are only results of underlying integration steps.

However, what has to be avoided is a missing


return value of a property function or a value which
cannot be handled by the solver. And this is what
IAPWS-IF97 disables for the perturbation analysis
itself. Here the benchmark revealed that
occasionally IAPWS-IF97 also returns non number
values (NaN) or infinity for some argument-tupels.

You might also like