ENERGY (2)

Copyright © IFAC PRP 4 Automation , Ghem, Belgium 1980

STUDY OF THE STEAM DISTRIBUTION

PROBLEM FOR A PLASTIC PLANT

P. A. Pilavakis* and L. de Wael**

-Process Design Department, ICI-Europa, Ltd., Everberg, Belgium

--Department of Applied Mathematics, State University of Ghent,
Ghent, Belgium

Abstract. This article describes a dynamic simulation of a steam distribution

system for a number of process plants to produce PVC plastics. The system
comprises of two fired boilers and six different levels of steam pressures.
There are also turbines, let-down stations, desuperheating units, deaerators,
safety relief valves and the associated control equipment. Several users are
supplied with steam of different pressure. Initially there are only two
plants I a VCM and a PVC. The VCM plant has rather large and sudden varia-
tions in steam demand. This makes a dynamic model particularly useful as a
check on the system. Some approaches for solving this stiff problem are tried
out. Solution on a digital computer allows a realistic model and accurate
results to be obtained but integration takes a long time unless sophisticated
methods or high-speed computers are used. On a hybrid computer integration is
accurate and results are obtained faster. However, the model has to be simpli-
fied by lack of analog elements and in non-linear function calculation
accuracy is lost. As a result of the simulation runs, useful modifications to
the original design are proposed and different "dangerous" operation strate-
gies are detected. It also allowed to obtain operational experience of the
system before start up.

Keywords. Boilers; control system analysis; dynamic modelling; hybrid

computers; nonlinear systems; process control; steam plants.

INTRODUCTION

The increased concern with the control of varying operating conditions. Furthermore,
steam distribution systems has led to a pro- the steam system has to be reliable since it
nounced interest in the dynamics of this is used as primary driving force for rotating
system. This is made practical with the pre- equipment in the central service area, for
sent availability of high-speed computers. instance, instrument air compressor, boiler
fuel oil and water supply.
A steam balance is calculated on the basis of
different operating steady states for steam Equipment utilized in steam systems varies
users (for examples see also Gordon and co- from plant to plant to such a degree that it
workers, 1978; Pilavakis and Perrin, 1980; is difficult to standardize on any single
Vogt and Wolters, 1976). A dynamic simulation steam dynamic simulation. In this article an
can be very useful to give results when tran- example of a steam distribution system is
sition between different operating steady chosen and worked out. The designer must
states are investigated (Stamper, 1973). Such therefore make adjustments to the diagram
a simulation allows the effect of transient presented here to suit his needs or develop
conditions on steam boilers, prevailing a similar diagram for his particular project.
pressures and flowrates in the distribution
system to be identified. It therefore allows
to readjust some parameters of the design DESCRIPTION OF THE STEAM
control systems, enables rapid stabilisation SySTEM
of the steam control systems during commis-
sioning and allows operational experience of The physical layout of the steam distribution
the systems to be obtained before start-up. system is Sh0wn diagramatically on Fig. 1.
Steam is generd.ted by two fired package
lni tially there are only two plants : p VCM boilers 1 and 2. The design capacity of each
and a PVC. The VCM plant has rather large and boiler is 100000 kg/hr. Boiler 1 is on auto-
sudden variations in steam demand. This makes matic pressure control while boiler 2 is set
a dynamic model particularly useful as a check at a fixed generation rate which for this
on the system reactions and stability under example is 30% of its design capacity (i.e.

395
 W
1.0
a-

~______________________~m4?
I

Po normal C 45 bar
HP VCM Header

VCM . HP. Users I m24

Flash IP S tf' Dl n '"0

Drum Gener'l tor »

m .'"0. .
r.l " 7 43 ......
III
~
P 7 normal .. 1(, bar ;>;"
.. .
IP VCM Header '"
III
;:l
Po.
r-'
m28 m35 VCM . lP. Users I m?6
o
ro
::E:
III
Flash lP Steam ro
......
Drum Generator

m39 111129 Im 36

P8 normal "" .., .~) bar

LP VCM Header

VCM . LP. Users m

30
Pi~. 1 Ste~ diotribution systom di ~-ram
 Study of th e Steam Distribution Probl em 39 7

its generation is 30000 kg/hr). Saturated The IP service area header on which during
steam in the boiler drums is produced at disturbances there is a lot of variation in
about 60 bar and 262.7°C. This is superheated pressure has also a controlled blow-off
to 400 C and introduced into the HP service valve to atmosphere (m S)' This prevents the
area header (mS' m ). The turbines of the 3
9 IP service area header safety relief valve
service area use HP steam and release either to lift during the event of raising IP
IP or LP steam. pressure. Safety relief valves should be
avoided to open, since once open they later
The IP service area header is supplied with tend to leak and therefore have to be
steam by release from the HP/IP turbines, by replaced.
the HP/IP let-down and by a condensate flash
drum (m , m , m ). The LP service area Finally, the deaerator LP steam supply valve
ll 4 19
header is supplied with steam by release from and the IP /LP let-down valve are monitored
by a split range controller. The deaerator
the HP/LP turbines, with steam supplied from
the blowdown flash drum, with steam supplied will be provided with LP steam whenever this
from a condensate flash drum and in case of is in excess (m ). The difference for the
6
steam shortage with steam supplied from the deaerator requirements are obtained from IP
IP/LP let-down (m lO ' m21 , m22 , m ). The steam (m 16 ).
5
deaerator is supplied with LP steam which is
in excess (m ) and with IP steam (m 16 ). Users
6 THE DYNAMI C MODEL
in the VCM plant are supplied from HP. VCM,
IP. VCM and LP. VCM steam headers (m , m ,
24 26 A steam distribution system is said to be at
m ). The HP. VCM header is supplied with steam unsteady state operation if one or more of
30 the variables of the system vary with time.
by let-down from the HP service area header
(m 12 ). The IP.VCM header is supplied with The analysis required to determine the
steam by let-down from the IP service area dynamic behaviour involves considerations of
header, with steam supplied from a condensate problems in mass transfer, unsteady state
flash drum and with IP steam generated in the material and heat balances and process
VCM plant (mlS' m , m ). The LP.VCM header control. Some assumptions and simplifications
25 27 built in the model are discussed. The first
is supplied with steam by release from the
step consists of setting up the mathematical
IP /LP. VClIl turbines, with steam by let-down
equations needed to describe the various
from the IP.VCM, with steam supplied from a
sections. The full details of the equations
condensate flash drum and with LP steam gene-
will be given.
rated in the VCM plant (m 2S ' m , m , m ).
35 39 29
The PVC plant is supplied with steam from the As it is impossible to obtain general
IP service area header (m ). analytical solutions to problems of this
31 type the desired solutions are found by use
of numerical methods. Different numerical
All let-down streams (except the IP /LP. VCM)
integration methods have been used and will
are desuperheated with boiler feed water
(m , m20 , m , m ). The exhaust steam from be discussed.
15 40 41
the HP/LP and HP /IP turbines is also desuper-
heated with boiler feed water (m , m ). ASSUMPTIONS AND
13 14 SIMPLIFI CATIONS
The HP service area header pressure is con-
As a compromise between complexity and
trolled by varying the firing rate to boiler
1 and in case of run away pressures by opening reality the following assumptions are made
the boiler safety relief valves to atmosphere
1. The distributed capacity of the steam
(m l , m )· headers is represented by lumped volumes with
33
uniform pressure. Any second order effects
All let-downs are downstream-pressure control- along the mains are neglected.
led. The HP/IP service area let-down is also
equipped with an override control system. In 2. The steam flow to the HP main from the
case the HP pressure drops below a certain boilers is calculated from the pressure
level it closes the HP /IP let-down valve and difference between the main and the boiler
prevents the HP pressure to drop further (this steam drums. A square root law for pressure
is a required protection essential to prevent drop along the pipes is used.
turbine drives to trip). To be protected from
run away pressures there are safety relief 3. Header pressure variations are simulated
valves on all six headers which will blow with the ideal gas law. The compressibility
excess steam to atmosphere under abnormal factor of steam in the headers is also
conditions (m , m , m , m , m , m , m ). included and is treated as constant during
l 33 2 3 42 43 36
variation of pressure. This gives a linear
However the HP service area header is equipped relationship between flows and the pressure
with two safety relief valves which also
in the headers.
provide individual protection to each boiler.
398 P . A. Pilavak i s and L. De Wael

4. The pipes in the distribution system are

well lagged and temperature changes can be
neglected.
Xi = Kie i + l/TI i J: Kieidt (2 )

where, Ki = G/TRi
S. VCM steam headers are included in the
simulation while PVC steam headers are not. Gi I controller gain
TR . I transmitter range (bar)
1
6. Both boilers are identical.
TIi I integration time constant (s).
7. The water in the boiler drums and tubes is Controller outputs Xi are limited between 0
well mixed and is in thermal equilibrium with
and 1.
the steam space and the metal of the drum and
tubes. All these valves are treated as equal
percentage valves and the fractional opening
8. The mass of water in the boiler drum is
f(y . ) is given by I
taken constant. 1

9. 'Ibe steam drum pressure change is derived f(Y.)

1
= exp(-A.(l
1
- y.))
1
from a simplified enthalpy balance.

10. Boiler 1 is set on automatic pressure or l. = 0 when y.l. ~ 0.001

f(Y.) ~

control. Boiler 2 is set at 30% of its design

capaoity. Boiler 1 maximum steam production (since otherwise f(Y . ) mathematically would
1
is 100000 kg/hr. Boiler 2 steam production is never be zero)
30000 kg/hr. where, Yi I valve stem fractional position
11. The turbines steam consumption is constant Ai : constant
and not a function of the steam produced by
the boilers or the pressure in the upstream The HP/IP service area let-down valve 4 is
or downstream headers. also equipped with an override control system.
In case the HP pressure (p ) drops below a
12. The deaerator steam consumption is 3
constant and not a function of the steam certain level the error to this controller is
produced by the boilers. then given by I

13. Steam produced from blowdown and conden-

sate flash drums are oonstant and not a
funotion of the steam produced by the boilers. where, CS , I controller 4 set point for
4
14. All safety valves are treated as linear minimum HP header pressure (ber).
on/Off valves. As a result tne HP/IP let~own controller
reduces valve opening and prevents the HP
lS. The let~own valves, the valve to the pressure to drop further.
deaerator, the IP service area header blow-off
valve and the fuel valve to boiler 1 are lP/LP let-down valve (VS). The IP/LP let-
treated as equal percentage valves. down valve 5 and the deaerator LP steam
supply valve 6 are split range controlled by
16. The IP service area header blow-off valve oontroller 7. The error e to this control-
controller is treated to be equipped with a 7
ler is given by Eq. (1) while the controller
device to prevent saturation of the integral output x is given by Eq. (2). The output to
term. 7
the deaerator LP steam supply valve 6 is
given by I
EQUATIONS
X6 = (0.8 - ~)/0.8 (S)
The model is described by the following
mathematical equations (see also Fig. 1) I
while the output to the IP /LP let~own
valve S is given by :
Valves

Let~own valves (CV4, CV9, CV13, CVl4). The

Xs = (~ - 0.8)/0.2 (6)
error to the let:aown controller 1S :
Safety relief valves (CVl z CV2, CV3, CVlO z
e.
1
(1 ) cvlI, CVI5, cV16). The error to the sat'ety
rell.er varve controllers is given by :
where, CS.1 I controller set point (bar)
Ph I downstream header pressure (bar)
The oontroller output is : where, Ph : header pressure (bar)
These are on/Off valves and the controller
 St udy of t he Steam Distri buti on Prob l em 399

output is given by : All valve position derivatives are given by

the following equations ,
x. 0 when e. <0
y'l <_
1. 1.
(8 ) when x1.. - 1.
1
or x. I when e ~ 0
1. i TC.
1. I TC.
1.
ST .
1.

These valves are treated as linear valves and

the fractional opening f(Y . ) is given by I dyi -1 x. - y . -1
1.
1. -<=- when 1.<_ (15)
dt ST . TC. ST.
1. 1. 1.

or dyi 1 x . - y. 1
Control valves 1 and 11 function also as
relief valves to the boilers. In case the
-
dt
=ST-. when 1.

TC.
1.>_
ST.
(16)
1. 1. 1.
boiler pressure rises above the boiler set
point the error to the corresponding control-
ler is given by : where, TC. valve time constant (s)
1.
ST. I valve stroking time (s)
1.
(10)
Flows
where, P I boiler pressure (bar)
b
I controller set point for
For steam critical flow (m , m2 , m , m , m ,
i
CS l 3 4 5
maximum boiler pressure (bar)
m , m , m , m , m , m ) through valves
33 35 36 38 42 43
The controller output and the fractional (i.e. l1P>-- 1/2 P ) the following equation is
u
opening of the valve is then calculated as
previously. used I

IP header blow-off valve (CVl2). Control (17)

valve 12 (see Fig. 1) l.S equl.pped with a
device to prevent it from saturating. The where, Ki = max(mi)/pu normal
error to the controller is given by Eq. (7).
The controller output is , P : header upstream pressure (bar)
u

~2 = G12e 12/TRJ2 For steam subcritical flow (m 6 , m12 , m18)

through valves (i.e. l1P < 1/2 Pu ) the
+ 1/TI 12 f~G12(eliTR12 - e )dt following equation is used I

(11 )
~
2 2 '
m. :: K . f(y . ) P - P (18)
1. 1. J U d
where e is given by :

e <= 0 when 0 < ~2 < 1

(during unsteady state)
t; = ~2 when 0 ~ x 12 (12 )

or e = ~2 - 1 when x 12 >- 1 where, Ki <= max(mi)/ ~p~ normal - p! normal

The reduced error input to the controller P d ' header downstream pressure (bar)
(e 12 /TR 12 ) is limited between -1 and +1. The
valve is treated as an equal percentage valve Water flows (m , m , m , m , m , m )
13 14 15 20 40 41
and is described by Eq. U). to desuperheated let~own streams are taken
proportional to the let~own steam flows.
Boiler fuel oil valve (cv8). The error to the
bOl.ler ruel 01.1 valve l.S gl.ven by I The steam flows (m , m ) from the boilers
8 9
to the HP header are calculated from
(13) equation:

where, P 3 : HP header pressure (bar)

The controller output and the valve fractional
opening are given by llIs. ( 2 ) and (3)
respectively.

Valve Positions (during unsteady state)

400 P . A. Pi l ava kis and L. De Wa el

where, P , P : pressure in the boiler and where, Mw' Mm : mass of water and of metal
b 3
the HP header respectively (bar) respectively (kg )
~ , ~ : specific heat of water
The following steam flows are taken constant "I m
Steam flows (m lO ' mll) to service area HP / LP and met a l respectively (J!KgOC)
dh/ dP : derivative of boiling "/ater
and HP/IP turbines. Steam flolVs (m , m ,
19 21 enthalpy to pressure around 50 bar (J/kg
m ) from the service area condensate flash ~ar)
22
drums and the blowdown flash drum to the IP C = 1000 zRT/(MH V. )
o
and LP service area headers. Steam flows where, z : compressibility factor gf
(m32' m , m ) due to losses in the service superheated steam at 50 bar and 400 C
33 34 R gas law constant (bar
area HP, IP and LP headers and flow due to
IP service area users. Steam flm-ls (m , m 6'
2 m3/mol K)
24
T I temperature of superheated
m , m28' m , m , m ) corresponding to
27 29 30 31 steam (K)
users or generators in the VCM or PVC plant. MW molecular weight of water
(g/mol)
To model the VCM headers in more detail the Vb gas volume in the boiler
steam flows (m , m ) from the IP and LP
25 39 drum (m 3 )
condensate flash drums in the VCM area are
taken proportional to 10% of the HP and IP
Headers
user flows (m , m 6) respectively.
24 2
The following equation simulates the
Boilers headers I

The following equations simulate the boilers dP . r.

-=-=K.Lm.
~j=l J
(fori 3t0 8 )
dP l ~ - m8 (h s - hw) dt
(20)
dt e dh/dP + (h s - h w )/C
}ij
where, Pi : header pressure (bar)
n number of flows entering or
dP 2 Q2 - m (h s - hw) leaving the header
9 (21 )
Ki = 1000 zRT/(r,il1 Vh )
dt Medh/dP + (h s - hw)/C
where, z : compressibility factor of steam
at the pressure and temperature of the
where, PI' P 2 : pressure in boiler 1 and 2 header
respectively (bar) T temperature of the header (K)
Ql' Q2 : proportion of heat supplied
Vh volume of the header (m 3 )
by the fuel to each boiler (J/s) given by the
equations:
THE COMPUTER PROORAMME
(22 )
The computer programme is written in
(for boiler 2 set at Fortran. A main programme defines the
constants and the initial conditions and
calls a subroutine for integration. The
307~ of design capacity) (23) latter requires another subroutine to
calculate the derivatives which describe
where, B : maximum heat supplied by the fuel the steam system.
to ea~~ boiler (J/s)
f(Y8) : boiler fuel oil valve Parameters of the system and integration
fractional opening calculated by Eq. (3). step length are read in by the main
m , m9 I steam flows from boilers 1 programme. Flow of users are provided by a
8
separate subroutine that a llows to apply as
and 2 respectively (kg/s)
a function of time disturbances of different
hs : enthalpy of superhe ated steam at types.
0
50 bar and 400 C (J/kg)
h'-I : enthalpy of boiler feed water at NUMERICAL SOLUTION
0
130 C (Jjkg)
Me I equivalent mass of water in each A general layout of the control system is
given in Fig . 2. Pressures are compared
boiler (kg) calculated by :
with their set points to give the errors.
These are inputs to the controllers which
M
e affect the valves. The latter act upon the
system to change the flows which are also
 S tud y o f t he S t eam Di s tribution Pr obl em 401

Input flows
Controller Valve
,---------'L
Errors
Controller
,----------,
Valve
,-----------,
Outputs
Valve F
Openings
Boiler and
Dynamics Dynamics Actions Header
x.l. Dynamics i
L -_ _ _ _ _ _ _ _--'

Fig. 2 Block diagram of the control system

influenced by the pressures. The flows was used. This routine is accurate but
determine the pressures in the headers and requires four function evaluations for each
the boilers. Pressures are also influenced integration which take a long time to
by input flows (changi~ steam demands and compute. Then a predictor-corrector formula
supplies to the headers). by Adams-Moulton (Lambert, 1973) which
requires only two function evaluations per
In the control loop there are three distinct step length was tried out. To obtain how-
blocks. Firstly the controllers secondly the ever the same accuracy a smaller step length
valves and thirdly the boilers and headers. was necessary and so there was no gain in
In this example the three blocks have time. Furthermore, the method did not work
different time constants. The controller so well because of the bound character of
outputs given by Eg. (2) have a time constant some variables.
TI. taken equal to 36 s. The valves have a
l.
To speed up the solution of the problem the
time constant TC of 0.5 s as shown from
i system was divided in parts with different
Eg. (14) and hence react faster. Non-linear time constants and was then integrated
equations for the headers are obtained if separately using different step lengths. As
the flow Egs. (17), (18) and (19) are sub- shown in Fig. 2 the valve positions do not
stituted into the header pressure Eg. (25). depend explicitly on the value of the
The time constants for these equations vary pressures and hence could be considered
with the value of the pressures and the valve constant during integration of the pressures.
openings from about 3 to 40 s. In a similar way the controller outputs could
be considered constant during integration
As time constants in these equations lie in of the valve positions. Therefore, the
a wide range we have a stiff system. The valve equations could eventually be inte-
first possibility of numerical solution is grated with a different step length than
to integrate all derivatives together in the pressure equations. Using the same
one integration routine. In order to obtain method as explained previously to derive
accuracy and stability of the results the the step length required, it was found that
step size of integration should then be with this method the required step lengths
smaller than the smallest time constant of for the two integrations was the same and
the system. This requires a long computing computing time remained almost the same. A
time. re a son for this is probably that the feed-
back on the system influenced pressures and
In the present example several step lengths valve positions during integration.
were tried out. The step length was halfed
until the re suI ts did not change and the Hamming predictor-corrector methods with
smaller value was chosen. It was found that variable step length was also tried out.
a step length of 0.25 s was required. The This algorithm halves the step length when
4th order ~~nge-Kutta integr ation a l g orithm the initial step length is chosen large and
avoids computational instability. However
when the small time constants do not
influence results it does not increase the
time information step length sufficiently in order to obtain
Boiler and Control a considerable saving in runtime.
Header and Valve
Dynamics Dynamics A more sophisticated method seems to be
Gear's method for stiff differential
equations. However no programme neither a
Analog valve openings Digital description of this method was available and
Computer Computer this was not tried out.
Fig. 3 Hybrid configuration Finally a hybrid computer configuration was
tried out. The boilers and headers were
40 2 P . A: Pilava kis and L. De Wae l

simulated on the analog computer while the carried out is presented I

oontrollers and the valves were simulated on
the digi tal compute r • The time consumi ng 1. The steam demand of the HP. VCM users
function evaluation of the pressure (m ) was changed from 20000 to 0 kgjhr
derivatives were calculated on the analog 24
(see Fig. 1). The effect of the gain of
computer. The pressures were converted to CVl3 on the system was investigated. It was
digital values by an ADC and read into the found that if the gain was too low safety
digital computer. These together with the
relief valve 15 opened. This was undesirable
controller and valve equations resulted in
and confirmed that the gain of control valve
digital values for the valve openings that
13 had to be high.
set the DACs in the analog computer (see
Fig. 3). In the run shown in Figs. 5 and 6 the gain
was high enough. The pressure of the YCM.HP
A hybrid solution is interesting in the
header (p ) raised but did not reach the set
initial stage of the simulation since it 6
provides an aoquaintance with the problem. point of the safety relief valve, it then
Changes in steam demand are simulated by remained constant. The flow through the let-
potentiometer adjustments and quick down valve (m ) dropped rapidly as the
12
graphical results of the pressure profiles
valve closed. As the pressure in the HP
are obtained. The control part of the system
can then be adjusted in the digital part of servioe area header (P3) increased the flow
the hybrid. from boiler 2 dropped (m ) and then oscil-
9
However, a hybrid solution presents some lated bac..l< to its fixed value of 30000 kg/hr.
drawbacks. The problem complexity is soon The flow from boiler 1 (m ) which was on
8
limi ted by the number of calculating analog automatic pressure control dropped. This
elements. There has to be scaling, patching indicated how boilers smoothed fluctuations
and checking of the circuit which is time of the pressure in the HP service area
consuming. Accuracy is also lost due to non- header by condensation and flashing effects.
linear computing elements.
2. The steam demand of the LP. YCM users
From all the methods tried out the digital (m ) was changed from 30000 to 40000 kg/hr.
solution with the Runge-Kutta integration 30
From this run it was shown (see Fig. 4) how
gives the best results and was used for the
the disturbance in the first LP.VCM header
final runs.
influenced the pressure in this header and
how this disturbance travelled with time
from header to header (from P to P to P
RESULTS AND DISCUSSION 8 7 4
to P ). Pressures in the headers fluctuated
lni tially runs were carried out to adjust 3
valve settings and gains. For this purpose within acceptable levels.
disturbances were applied on each header one
at a time. When all valve parameters were 3. The steam supply from the IP. YCM genera-
adjusted further disturbanoes vlere implemen- tor (m ) was changed from 39200 to 0 kg/hr.
ted and several runs were carried out. These 27
disturbanoes were made stepwise and simulated Pressures in the IP. YCM and the IP servioe
area headers dropped considerably and a
the worst conditions sinoe in practice
changes are smoother. A selection of the runs recommendation will be made to the plant,
in case of voluntary stop of the IP.YCM
generator steam supply, to implement this
change gradually.
minimum 4. All HP /IP sel'vioe area turbines were
pressure"-yl P tripped. Turbines steam demand (m ll ) was
50 3
then reduced from 4 3000 to 0 kg/hr. The
P results are shown in Figs. 7 and 8. At the
20 4
moment the turbines were tripped there was
Pressure
a small incre ase in the HP header pressure
Ch:) 14.8 (p ) and a fall in the IP he ader pressure
(bar 3

P8
I 1 bar
(p 4). As the HP/IP let-dOl"ffi valve 4 reacted
fast, the let-dovm flo'~ (m ) incre ased
4
rapidly and the pressures returned to their
4.5 initial value. When turbines were tripped
there was a ~~-.c~tary ste8~ surplus in the
0 20 40 60 HP header and the flOl-ls from the boilers
Time (s) (m , m ) dropped sharply .
8 9
Fig. 4 Run 2
5. The steam demand of the lP.PVC users
(m ) was changed from 34000 to 0 kg/hr.
31
 Study of the Steam Distribution Problem 403

50 p) ml!
-'
me
40
Pt:
48
m9
Pressure Flow 30
(bar ) 10- 3
4G (kgjhr)

20

44
0 20 40 60
Time (5)
Fig . ", Run 1 10
-
0 20 40 60
Time (5)
Pig. 8 Run 4
40

!'!le
50 ?3

Flow
10- 3
(kg/hr) 20 23
Pressure
change 22
(bar)
2l
10
1::12
20
I I bar

0 20 40 60
0 Time (5)
0 20 40 60
Time (5) Fil?' . 9 Run c:

Pi 6-- - 6 Run 1

4,:",

50 ~Q
F"(
~ c;

Flaw
i'reesure 10- 3 .........
~

20 (~:g/r.r )
30 -'
ch:l.'1,:'
(bar )
19
7. 5
18; c:

0
c.: 20 .10 60 Co 40 60
Time (5) Time (5 )
Pi ~ . 7 Rt:n t. Fi? . 10 ?" U1 "
404 ~. A. Pilavakis and L. De Wael

It was found that the blow-off valve had to be standard blocks (e.g. boilers, turbines,
equipped with a device to prevent the integral headers, users).
part of the controller output from saturating.
If this was not the case and the valve
remained closed for certain time even if the ACKNOWLEOOMENT
pressure in the IF header raised above the set
point this valve would not open directly and He would like to thank Professor
safety relief valve 2 would open. The same G. C. Vansteenkiste vrho kindly provided the
would happen if the valve remained open for computer facilities in his department at
a certain time and later pressure in the IF the University of Ghent, where this study
header dropped below the set point of the was carried out.
blow-off valve the latter "Would not close
directly. Fq. (ll) describes an example Colleagues within ICI, particularly
formula to prevent this valve from saturating. J. Anderson, M. Oglefby and A. Scott, are
acknovlledged for their help during develop-
It was also found that the gain of the IF ment of this work.
blow-off valve had to be high. With a small
gain, blow-off through this valve was small, v~ would also like to thank the management
pressure in the IF header increased and the of ICI-Europa for permission to publish this
IF safety relief valve 2 opened. article.

The size of the blow-off pipe was also

initially designed small and the IF safety REFERENCES
relief valve was bound to open. Recommenda-
tions were therefore made to make this pipe Gordon, E., M. H. Hashemi, R. D. Dodge and
larger. J. La Rosa (1978). A versatile steam
balance program. Chem. Eng. Prog., 74,
Finally, since under these conditions of steam 51-56. --
requirement there was an excess of IF steam Lambert, J., D. (1973). Computational Methods
which ,"las blown-off to the atmosphere it was in Ordinary Differential EqUat~ons.
recommended to transfer some of the HP/IF John W~ley &Sons, London. p. 85.
turbines to electrical driven motors. Pilavakis, P. A. and 111. Perrin (1980).
Steady state simulation of a steam
When all recommended modifications were distribution system. To be published.
implemented, results were obtained as shown Stamper, H. (1973). Analogue computer
in Figs. 9 and 10. The IF pressure (p 4) simulation of a methanol plant steam
system. Symposium Simulating Boilers
increased to an acceptable level and then and Steam D~str~bUt~on Systems.
dropped since the HP/IF let-d.own valve (flow
m ) closed and the control blow-off valve 12 Vogt, G. A. and M. J. Wolters (1976). Steam
4 balances as a working tool. Chem. Eng.
opened (flow not shown on the graph). Prog. , 72, 62-65.

6. Boiler 2 was tripped from steam supply

of 30000 to 0 kg/hr. Boiler 1 steam production
increased gradually to the initial level of
steam produced by the two boilers. Pressures
wi thin the HP and IF service area headers
dropped considerably. A recommendation will
be made to the plant, in case of voluntary
stop of boiler 2, to implement this change
gradually.

CONCLU5IONS

Results obtained from the dynamic model

provided guidelines for controller settings
and gains, led to usefull modifications to the
original design of the control system and
provided recommendations for plant operation.

A fast computer or sophisticated methods,

programmes and algorithms are required to
simulate the dynamic behaviour of the system
over a long period of time.

For different steam systems, it is rec:uired

to rewrite the programme. It could therefore
be useful to develop a general purpose
programme which could be applied to different
configurations by definition of a number of