You are on page 1of 8

157

Transient analysis of power plant ,deaerators


R D Bell, BE, PhD, MIEEE
School of Mathematics,Physics, Computing and Electronics, Macquarie University, Sydney, Australia
N W Rees, BSc, PhD, FIEAust, SMIEEE and c Lu, BSc, MSEE, MIEEE x
Department of Systems and Control, IJniversity of New South Wales, Sydney, Australia

A mathematical model suitable for predicting the transient behaviour ofdeaerator plant/Es presented. The model is based on energy and
mass balance equations applied to the deaerator, storage tank and feed p t m p suction pipe. The model parameters are based entirely on
the physical characteristics of the plant, and hence the model can be easily adapted to plants of diflerent size. A comparison with data
collected from a deaerator plant operating on a 500 MW unit indicates that the model gives good transient responses and can be used
to predict accurately the onset of cavitation in the feed pump.

NOTATION Recirculation loops


Deaerator and storage tank t,, time of water travelling round deaerator circu-
specific enthalpy of condensate (kJ/kg) lation loop (s)
specific enthalpy of make-up water (kJ/kg) t,, time of water travelling round feed pump circu-
specific enthalpy of steam (kJ/kg) lation loop (s)
specific enthalpy of water (kJ/kg) wCircl deaerator/storage tank water circulation mass
water level in deaerator storage tank (m) flow (kg/s)
length of storage tank (m) wCircZ water circulation mass flow for feed pump
total mass of water and steam (kg) output loop (kg/s)
mass of steam (kg)
mass of water (kg) 1 INTRODUCTION
steam pressure in deaerator (kPa) The model derived in this paper can be used to predict
radius of storage tank (m) deaerator plant behaviour under transient conditions. It
specific internal energy of mixture (kJ/kg) is a particularly simple model, consisting of three differ-
specific internal energy of steam (kJ/kg) ential equations and a small set of algebraic equations
specific internal energy of water (kJ/kg) that represent the steam tables. The three differential
specific volume of steam (m3/kg) equations are coupled non-linear differential equations,
specific volume of water (m3/kg) and although this may seem to complicate the model
total volume of deaerator and storage tank (m3) structure, it overcomes the problems associated with the
condensate mass flowrate (kg/s) conventional method of using an iterative procedure to
make-up water mass flowrate (kg/s) solve a set of non-linear algebraic equations for this
mass flowrate of fluid into suction pipe (kg/s) type of problem. This approach has resulted in a model
deviation from tank centre-line (m) that seems to predict plant behaviour under transient
conditions better than other models that have appeared
in the literature (1-3).
The model is derived from mass and energy balance
Suction pipe
equations applied to lumped sections of the plant. The
specific enthalpy of water at feed pump inlet coefficients in these balance equations come from the
(kJ/kg) physical dimensions of the plant, and hence the model is
head equivalent of vapour pressure in deaerator easily adapted to different size plant.
(4 The model is only valid up to the point of feed pump
head equivalent of vapour pressure of water at cavitation. No attempt at this stage has been made to
feed pump (m) model the cavitation phenomena, as it is felt that, if a
head equivalent for pressure drop due to fric- correct control strategy is implemented based on this
tion in suction pipe (m) model, then the plant would not reach that condition.
head of water in suction pipe (m) Mechanical damage results quickly after the advent of
mass of water in suction pipe (kg) cavitation, and this region of operation should be
vapour pressure at feed pump (kPa) avoided.
total pressure at feed pump (kPa) To gain greater confidence in the model, a compari-
time delay of water through suction pipe(s) son with data collected from a deaerator operating on a
water flow through feed pump (kg/s) 500 MW unit is given. This shows that the pressure
transient in the deaerator plant is modelled well, and
The MS was received on 20 February 1992 und was accepted for publication on the prediction of the onset of cavitation is quite accu-
6 August 1992. rate.
I00592 Q IMechE 1992 095Y-6518jY2 53.00 + .05 Proc Instn Mech Engrs Vol 206

Downloaded from pii.sagepub.com at PENNSYLVANIA STATE UNIV on September 17, 2016


158 R D BELL, N W REES AND C X LU

2 MODEL DEVELOPMENT set of initial conditions and then use an iterative non-
The structure of the deaerator plant being considered is linear algebraic procedure to find the masses m, and
illustrated in Fig. 1, which shows two inputs. One input m, . This is possible since, under saturated conditions,
is condensate water w, with enthalpy h, that might be there is a non-linear relationship between the pressure,
varying, and the other is make up water we from a fixed the specific enthalpies, specific internal energies and
tank with constant water enthalpy he. Extraction steam specific volumes. The procedure involves using the
flow from the turbine section is not shown as an input constant volume equation and steam tables (or a math-
because the model is concerned with the transient ematical approximation for them) for a saturated fluid.
behaviour after a plant trip, when this Bow is zero. It The convergence and computational overhead of this
would, however, be a trivial matter to add it as an procedure has resulted in little use being made of it in
input. practice. The general approach has been to assume
good water level control, and hence the mass of water
remains constant or varies slowly with time. However, if
2.1 Deaerator and storage tank this simplification is made, poor dynamic performance
The deaerator and storage tank are considered as one during transient conditions can occur and the predic-
unit, and the energy balance equation for the steam and tion of feed pump cavitation becomes inadequate. This
water mixture in this volume with no recirculation can is further discussed in Section 4.
be written as To overcome this problem, a solution based on
solving the coupled differential equations is proposed.
This is similar to the approach taken in Astrom and
Bell (4) for simulating drum boiler plant.
Since the volume I/ is fixed, then
mlul = m,u, + m,u, (2)
The mass balance equation for the material in the
dV - d(v, m,
_ -
+ u, m,) =O
dt dt
deaerator and storage tank is
and, because of the saturated state of the fluid, this
d(m, + m,) = we + w, - equation can be expanded to
dt
If it is assumed that the steam and water mixture in the (7)
deaerator and storage tank is in a saturated state, then
equations (1) and (2) give Equations (3), (4) and (7) form three coupled non-linear
differential equations:
dm, dm,
U, - + u, - +
dt dt -dm,
+ - - - =dmw
d,
dt dt
= wehe(4) + w,h, - w,h,
dm, dm, dP1
A further equation is required to solve for the three
unknowns pl, m,, and m,. Since the deaerator and
a, dt + a2 -dt
+ a3 -dt
= d, (9)
storage tank are of a known fixed volume, dm, dm, dp
V = u,m, + v,m, (5)
b1-+
dt
b, -+
dt
b3 A =
dt
0

One method of solving this class of problem has been to where


step the differential equations (1) and (3) from a given
u, = u s bl = V,
wc wc
I I
a2 = u, b2 = V,
6V
b, = (m, + m, -

dl = W, he + W , h, - W , h, dz = W, + W, - W,
Since the water and steam mixture is in a saturated
state, h,, u,, u s , u,, u s , 6uS/6pi, 6uwI6pi, ~ V , I J Pand
~,
6vW/6p, can be expressed as a function of the vapour
pressure in the tank. This function can be either a table
look-up of the steam tables or, as has been used in the
following simulation, a linear or quadratic function
fitted to the steam tables for the normal operating con-
ditions of the process as outlined by Astrom and Bell
(4).
Given the initial conditions for pl, m,, m, and the
input and output functions, then the linear algebraic
equations (8), (9), and (10) can be solved for the
'~'Wl
unknowns dpJdt, dm,/dt, and dm,ldt. These differen-
Fig. 1 Schematic diagram of deaerator plant tials can thcn be stepped by an appropriate integration
Part I: Journal of Systems and Control Engineering @ IMechE 1992

Downloaded from pii.sagepub.com at PENNSYLVANIA STATE UNIV on September 17, 2016


TRANSIENT ANALYSIS OF POWER PLANT DEAERATORS 159

routine to give pl, m,, and m, . The procedure is repeat- 2.4 Recirculation loops
ed to obtain the time trajectories for pl, m,, and m,. Two recirculation loops have also been included in the
This is further described in Section 4. model. These loops are present in most deaerator plants
for pump protection and have been included here to
investigate their effect in controlling the NPSH.
2.2 Storage tank level and steam bubble formation The first loop is around the storage tank and the
deaerator. It essentially pumps the water from the
Once m, is known, the storage tank water level can be bottom of the storage tank and mixes it with the con-
calculated as outlined in the Appendix. To add a further densate before returning it to the deaerator. The right-
degree of reality to the water level signal, the steam hand side of equation (1) is changed to
bubble formation below the water level is modelled by
including a term proportional to the change of steam wehe + w c lac - w o h w - W c i r c l h w + W c i r c l h w ( t - fd (15)
production. The steam bubbles, after formation, travel This change will also occur in d , in equation (9). Note
quickly to the surface of the water, where they are re- that no change is necessary in the mass balance equa-
leased into the steam section. This formation, and the tion, since the mass flow out is equal to the flow in.
subsequent release from the surface, cause a large turbu- The other recirculation loop is to maintain feedwater
lence to the water level. This process has been captured pumping capacity under low feedwater conditions to
by including a random number with rectangular dis- the boiler. It occurs after the feedwater pump and recir-
tribution in the interval [0, 11 which modulates the culates back to the storage tank. For the condition with
steam rate term. The range of this distribution is posi- both recirculation loops active, the right-hand side of
tive to reflect the physical phenomenon of bubbles equation (1) and dl become,
forming below the surface, thereby only affecting the
water level in a positive direction. So the final storage we + wc hc - wo hw - wcirclhw + Wcirclh,(t - td2)
tank water level equation becomes
+ WcircZ h,(t - td3) (I6)
dm
H = x - kz rect(t) + r
dt 3 EQUILIBRIUM AND INITIAL CONDITIONS
where k , = 0.10 has been fitted to the data to give an Equations (8) and (9) can be used to calculate the equi-
appropriate variation in the water level during the tran- librium conditions when dp,/dt, dm,/dt, and dm,/dt are
sient. equated to zero. This gives
w, = we + w, (17)
2.3 Suction pipe
The suction pipe is assumed only to have water present,
and so this model is valid only up to the point of steam If a specified initial water level is given, m, can be calcu-
generation. Since the density of water is assumed to be lated from the dimensions of the storage tank, since the
constant for the operating conditions, the mass balance specific density of water for the normal operating of the
equation becomes an algebraic equation giving process can be assumed constant at 909 kg/m3. Theo-
retically the vapour pressure in the tank can now be
W0"t =wo (12) calculated, since the volume occupied by the steam is
The energy balance equation under this steady flow known and the steam tables relate the enthalpy h, to
condition is then this pressure. In practice, it is easier to specify a steam
pressure p1 in the tank and use this in conjunction with
'2 = hw(t - l d l ) f dHsp + H, (13) the mass of water m, (from water level) to calculate the
where t d 1 is defined in the notation and is a function of mass m, of steam. The enthalpy la, can now be calcu-
mass flow, and g is the gravitational constant. lated from steam tables ; however, this will not be the
Equation (13) gives the enthalpy of the water at the equilibrium condition, and so a transient response will
bottom of the suction pipe. h,(t - tdl) is therefore the occur.
enthalpy of the water at the bottom of the suction pipe For the suction pipe, the equilibrium condition for h,
without the combined effect of the head of water in the is given by equation (1 3).
suction pipe and in the storage tank. If the vapour pres-
sure equivalent to this enthalpy becomes greater than 4 SIMULATION RESULTS
the total pressure at the bottom of the suction pipe,
steam will start to be produced and pump cavitation 4.1 Simulation method
will occur. In fact, cavitation will occur slightly before The model has been simulated using two different simu-
this point because the pump has to have a net positive lation packages. To get a quick feel for the validity of
suction head (NPSH) for proper operation. Karassik (5) the model, the SIMNON package (6) was used. To
defines the NPSH as, permit easy investigation of more extensive scenarios
NPSH = H,, + H + Heqpl - HeqpZ- H , (14)
and also to allow a greater degree of real-time inter-
action, the model was also run on the University of
To avoid cavitation the NPSH must be greater than a NSW simulator shell (7).
value recommended by the pump manufacturer The simulation process is as follows. Using the given
(NPSHR). initial condition for pl, the steam table polynomial fits
0 IMechE 1992 Proc Instn Mech Engrs Vol 206

Downloaded from pii.sagepub.com at PENNSYLVANIA STATE UNIV on September 17, 2016


160 R D BELL, N W REES AND C X LU

are used to determine u,, u s , v,, us, h,, 6uS/6p,, 4.2 Simulation scenarios
6u,/6pl, 6vJ6pl, and 6vW/6Pi.These together To show that the new model has better transient per-
with the independent inputs w, , h, , w e , h e , wUut, wCircl formance than that of other models in the literature (2,
and wCirc2 are then used to
I the equa-
tions (8), (9) and (10) to determine the differentials 3, g), models have been tested on the following scenario:
A unit trip has occurred and no extraction steam is
dm,/dt, dm,/dt and dpddt. The differentials are then passing to the &aerator. Feed to the boiler is at 25 kg/s,
integrated to give m ~ m,w and P1. h2 is from and the condensate flow and make-up water flow are
equation (13)' The process is repeated at every integra-
zero. A stable condition is reached with the water level
tion step. The water level and the NPSH can be near its normal operation level and just slowly chang-
determined from equations (1 1) and (14).

(a)
PI
0- lo00 kPa

mw
0-300 OOO kg

H
2-4 m

m,
0-1W kg

We
0- loo0 kg

we
0-loo0 kgis

Wuut
NPSH
-
I""I""I1"~I~"~I"'~I""I"'~I""I""I""I"'~I'~'""''I

Part I : Journal of Systems and Control Engineering @ IMechE 1992


Downloaded from pii.sagepub.com at PENNSYLVANIA STATE UNIV on September 17, 2016
TRANSIENT ANALYSIS OF POWER PLANT DEAERATORS 161

ing. To restart the unit, feed flow is increased to 125 assumption is not valid and the consequence of this is
kg/s to purge the economizer. When the water level an unrealistically large change in pressure p1 following
drops to its low water level, the on/off controller the increased feed flow and a rapid drop in the NPSH
switches on the make-up water at 200 kg/s to the de- so that cavitation is predicted to occur at about 230 s,
aerator. which is well before the introduction of the condensate
Figure 2 shows the simulation results obtained using water. In reality, however, the plant would not cavitate
the literature models and the inputs from the above sce- in these circumstances, and this is confirmed by the new
nario. These models essentially assume that the water model in Fig. 3.
level is more or less constant, since they use an energy Figure 3 shows the results obtained using the new
balance equation of the form of equation (4) with dm,/dt model. Cavitation is not predicted, either following the
and dm,/dt zero. As can be seen from m, in Fig. 2, this increase in feed flow at 120 s or in the make-up water at

2-4 m
-

0-loo0 kg

wo

0 loo 200 300 500 600 700 800


Time

(b)
NPSH Pt \
0-30 m

PI
0- 1200 kPa

P2
0- 1200 kPa

NPSHR
0-30 m
\ P2

++',"I
0- 1000 kg

> , , I I , , , n o , I I I I I I I Z I b I I , 1 1 1 I I I V I I I I I I I I I I I I I I \ I I I I I I I I I I I 1 1

Q IMechE 1992 Proc Instn Mech Engrs Vol 206

Downloaded from pii.sagepub.com at PENNSYLVANIA STATE UNIV on September 17, 2016


162 R D BELL, N W REES AND C X LU

5 COMPARISON WITH PLANT DATA


Table 1 Main physical dimensions of
deaerator plant Table 1 outlines the main physical dimensions of the
Wallerawang
deaerator plant in a 500 MW unit of the Electricity
Commission of NSW at Wallerawang power station.
Length of deaerator 7.06 m A number of feedwater pump seizures have occurred
Radius of deaerator 1.25 m recently on this plant as described by Rovnak et al. (3).
Length of storage tank, 1 25.9 m
Radius of storage tank, r 1.9 m Using the above data and the new model, one such inci-
Water volume at N WL 250 m 3 dent has been set up and simulated as shown in Fig. 4,
Head of suction pipe, H,, 33 m where the actual plant measurements are also shown by
Radius of suction pipe 0.15 m the circles and crosses. The incident involves a unit trip
Friction head, I I , 5m
Enthalpy of condensate, h , 717 kJ/kg followed by a manual attempt to control the plant. At
Top tank steam pressure, p 1 1000 kPa the start of Fig. 4 the situation has been reached where
Too tank water temnerature 180°C there is no feed water to the boiler and condensate
water w , is being used to increase the storage tank level
which had become too low. The condensate valve then
315 s. In this model, the pressure p1 stays reasonably closes at 80 s, and at 120 s the feedwater pumps are
constant following the feedwater flow increase because restarted on a low flow. A reasonably stable deaerator
of the steam generation from the hot water in the tank. performance results, with the level only slowly dropping
The dropping water level, due to increased wo, causes and the NPSH very healthy and building up. At 360 s,
the make-up water to be switched on at about 315 s, the condensate valve is opened further to increase the
and this, as can be seen in Fig. 3b, causes the NPSH to storage tank water level. This results in a collapse of
drop quickly, but its value does not reach the NPSHR deaerator pressure and subsequent cavitation at about
value required before cavitation occurs. Therefore there 450 s. The figures show that the model predicts the
is no cavitation. Note, however, that the model does behaviour up to cavitation extremely well. After cavi-
show an increased agitation in the water level, Fig. 3a, tation, the conditions in the deaerator, Fig. 4a, continue
following the introduction of the make-up water. This is to be predicted well, suggesting that the local pump
largely caused by the steam generation associated with cavitation conditions are not reflected back into the
the influx of the make-up water and does not necessar- deaerator. By contrast, as might be expected, Fig. 4b
ily mean the start of cavitation as is sometimes thought fails to predict suction pressure at the feed pump. To do
by operators who hear the increased agitation. Note this, a cavitation model would be needed. The model
also that the vapour pressure p z at the pump inlet is predicts pressures significantly above measured pres-
always less than the total pressure pt at that point, sures, which reflects the fact that, on the plant, steam i s
which is to be expected if cavitation does not occur. being generated and passes through the pump into the
It is concluded from the above experiments and other feedwater system. The model also suggests that, after
less extreme experiments carried out with this model, cavitation, the vapour pressure p z holds quite well,
including tests checking recirculation loop effects, that reflecting that the suction pipe flow has been reduced,
the model is a much more realistic description of the and this slows down the heat transfer flow.
transient behaviour of a deaerator than the constant
mass models. This has been confirmed with plant engi-
neers. It is also noted that the model complexity is such 5.1 Data table for real datalsimulation comparison
that it can be readily simulated on an AT computer. In The plant data in Table 2 were sampled manually. The
the following section, the model is checked against gaps in the columns for the plant data indicate that no
actual plant data. data were recorded. Most sampling was carried out
from 360 s to 720s when the plant was seen as having
operational problems. The simulation data columns
Table 2 Plant and simulation data represent sampled values of the simulation results
shown in Fig. 4.
Time -
- *H -
H *PI -
PI *Pt
- -Pt
S m m kPa kPa kPa kPa
6 CONCLUSIONS
Plant Simulated Plant Simulated Plant Simulated
data data data data data data A simple model that seems to describe the behaviour of
0 2.0451 962.527 1227.31 deaerator plant under severe transient conditions has
60 1.8749 2.0645 817 863.552 1124 1128.51 been developed. The model extends those that have
75 1.9753 839.980 1104.15 been described to date in the literature by including the
120 2.0298 83 1.849 1096.50
I80 2.0109 833.340 1097.82
effects of steam mass variation and the formation of
240 2.0006 834.619 1099.01 steam bubbles. Simulation results from the model, when
300 1.9897 835.865 1110.16 compared with actual plant data, have given sufficient
360 2.1073 1.9749 810 837.107 1116 1101.27 confidence in the model for automatic control studies
385 2.0967 798.832 1064.07
420 2.0946 2.1988 732 715.473 1046
and optimum operating principles for deaerator plant
981.613
480 2.1113 2.2679 625 587.408 684 854.157 to be investigated.
540 2.3368 2.3592 480 475.484 555 743.038
600 2.6065 2.5075 379 377.315 44 1 646.177
660 2.8660 2.6117 355 290.643 437 560.423
ACKNOWLEDGEMENTS
720 2.7415 2.7631 332 213.629 636 484.742
780 2.9182 144.750
The authors wish to thank John Marshall and Bill
417.231
Peattie, of the Turbine and Cooling Water Section of
Part I: Journal of Systems and Control Engineering 0 IMechE 1992
Downloaded from pii.sagepub.com at PENNSYLVANIA STATE UNIV on September 17, 2016
TRANSIENT ANALYSIS OF POWER PLANT DEAERATORS 163

(a)

PI
0- 1200 kPa

H
1-3 m

WC
0-1000kg

W"
0-loo0 kg/\

0 100 200 300 300 500 600 700 800


Time

Fig. 4 (a) Real data: deaerator and storage tank


(b) Real data: suction pipe

the Electricity Commission of New South Wales, for REFERENCES


bringing this problem to their attention and for supply-
ing the data. The facilities used for the simulation have Thurston, R. s. Design of suction piping and deaerator storage
been developed over a number of years with the support capacity to protect feed pumps. Trans. ASME, J . Engngfor Power,
of the Electrical Research Board of Australia and the Ser. A, 1961,83, January, 69-73.
National E~~~~~ Research and Development pro- 2 Ray, A. Transients in power plant feed pump system. In Simulation
Council Proceedings Series: Simulation qf Energy Systems (Ed. K. E.
gramme. The authors would also like to thank the F. Watt) 1978,8, (2),December, 129-133.
~evkWersfor their stimulating comments which have 3 Rovnak, J. A., Wotring, T. L. and Marshall, J. Evaluation of NPSH
helped to improve this paper. during hot restarts of h s i l plants with deaerators. ASME Interna-
@ IMechE 1992 Proc Instn ,Mech Engrs Vol 206
Downloaded from pii.sagepub.com at PENNSYLVANIA STATE UNIV on September 17, 2016
164 R D BELL, N W REES AND C X LU

tional Joint Power Generation Conference, San Diego, Ca, 1991,


1-9.
Astrom, K. J. and Bell, R. D. Simple drum boiler models. IFAC I
Symposium on Power systems, Brussels, September 1988, 4.2.1 -
4.2.5.
Karassik, I. J. Are you short of NPSH? Nine ways to improve
unfavourable suction conditions. Combustion, 1980, July, 37-41.
Elmqvist, H., Astrom, K. J. and Schonthal, T. SIMNON-users
guide, 1988 (Lund Institute of Technology, Department of Auto-
matic Control, Sweden).
Kees, N. W. Modular simulation systems. IFAC World Congress,
Fig. 5 Solving storage tank level x
Tallinn, Estonia, 1990, vol. 10, 116-121.
Liao, G. S. and Lenng, P. Analysis of feedwater pump suction pres- It is not a trivial matter, given v, to solve this equation
sure decay under instant turbine load rejection. Trans. A S M E , J . for x as is required in the simulation. One way of
Engngfor Power, Ser. A , 1972, April, 83-90. solving the problem is shown in Fig. 5. This is a feed-
back structure with the set Doint u,, remesenting the I

APPENDIX volume of liquid in the tank as calculated from m, using


equations (8) to (10). The feedback loop compares v d
Storage tank water level with the volume u calculated from a given water level x.
This Appendix is concerned with the determination of The error ud - u is then fed into an integrator to
water level in the storage tank. The tank is cylindrical in correct the level x that is used in the non-linear equa-
shape and the volume occupied by the water for a tion. Since, in the steady state, the integrator will only
level of x metres from the centre-line is given by stop correcting the level x when u and vd are equal, then
the steady state output x from this loop will give the
required solution to the non-linear equation. The
response of this loop has to be selected to be compatible

- { 1-- (x r ) ] / ( 2 (x r, ((" I)'}" (19)


with the response of the model. If it is too fast, insta-
bility will result, if it is too slow, the level x will lag
behind the true value.

Part I: Journal of Systems and Control Engineering @ IMechE 1992


Downloaded from pii.sagepub.com at PENNSYLVANIA STATE UNIV on September 17, 2016

You might also like