Modelling of a two-stage high-pressure gas

reduction station
I. Fletcher, C. S. Cox, W. J. B. Arden and A. Doonan

Control Systems Centre, SEAT, Urziuersityof Sunderland, Sunder-land, E&and

The delivery of gas from the national transmission system into the regional supergrids, at the correct pressure, is
accomplished using a facility referred to in the industry as an above-ground installation (AGI). Early attempts
to control such systems were compromised by inadequate models developed to explain system behauiour. This
paper outlines the construction of a multi-input multi-output system model which has been used successfully to
explain system operation to gas engineers as well as prociding a basis for the design and implementation of a
number of advanced control systems. 0 1996 by Elsevier Science Inc.

Keywords: gas transmission, pressure reduction systems, pipeline and valve modelling

1. Introduction lyst for the development of schemes to overcome the

current operational problems and thereby greatly improve
The work described within this paper develops a line of
the efficiency of control.* Here one particular section of
analysis which commenced in the mid-1980s, when re-
the transmission system is considered whose role is to
search was initiated at British Gas Engineering into sys-
supply gas to the regional authorities at required pres-
tem re-inforcement requirements to fulfill the increasing
sures, the above-ground installation (AGI). The AGIs are
demand for gas. As time progressed, this growth in de-
situated at various locations along the length of the na-
mand has slowed down significantly, resulting in a change
in emphasis from construction towards one of efficient tional transmission system pipeline and are configured to
operation.’ extract gas at high pressure for delivery to the regional
The delivery of gas from the national transmission to transmission systems at the required pressure and flowrate.
Figure 1 illustrates a typical site configuration where the
regional systems and ultimately to the consumer, requires
a number of control functions to be performed. The prime main regulating stream involves three regulators con-
objectives when performing these functions are: nected in series. The gas enters the stream through a
first-stage regulator, which is normally fully open and is
1. system safety, integrity, and security of supply; activated upon sensing an excess pressure at the station
2. minimizing operation costs; and finally outlet. This is normally referred to as a slamshut, which
3. maximizing system throughput and storage. when activated will provide complete isolation between
The performance of gas control systems is influenced by the input and the output of the station, resulting in the
restricted pressure regimes caused by mechanical stress protection of the downstream pipeline from exceeding
ratings. Up until the 1990s this problem was further design pressure limits. Immediately following the slamshut
exaggerated by redundancies which have to be built into is the first-stage pressure regulating valve, known as the
the system because of the poor accuracy achievable using monitor. This operates as a fully modulating regulator in
existing pneumatic hardware. In addition, recent increases closed-loop pressure control, which is employed to reduce
in demand and seasonal variations have compounded sta- the inlet pressure to some intermediate value between the
bility and interaction problems. station inlet and outlet limits. The outlet from the monitor
Modern control technology coupled with major parallel supplies a second regulator referred to as the active,
advances in microprocessor technology provided the cata- which implements a closed-loop control strategy to regu-
late either the outlet pressure or volumetric flow rate.
These three regulators form a single stream and the
Address reprint requests to Prof. C. S. Cox, Department of Engineer-
complete structure is called an Active Monitor and
ing and Advanced Technology, University of Sunderland, Chester Slamshut Strategy. This type of station suffers from the
Road, Sunderland SRl 35D England. disadvantage that, with certain designs, instability can
arise at specific flows and pressure.3 This occurs as a
Received 5 June 1995; revised 16 February 1996; accepted 23 April result of interaction between the regulators and connect-
1996 ing pipework and manifests itself as a rapidly fluctuating

Appl. Math. Modelling 1996, Vol. 20, October

0 1996 by Elsevier Science Inc. 0307-904X/96/$15.00
655 Avenue of the Americas, New York, NY 10010 PI1 s0307-904x(96)00071-6
Modelling of a two-stage high-pressure gas reduction station: I. Fletcher et al.

of solving them, can then be considered as a simulation of

the system allowing systematic investigation of control
modes or, indeed, the effects of varying any parameter
within the jurisdiction of the designer.
The second (Section 3) is associated with physical mod-
els where actual hardware is constructed according to
appropriate scaling laws such that its behaviour is pre-
dictably related to that of the full-scale system. Wind
tunnel testing of aircraft and towing tanks for ship design
are examples of this particular type of modelling.
PIale - “S YzgEY
Figure 1. General layout of an AGI station.

2. AGI mathematical model development

interstage pressure, giving rise to rapid wear of the first-
stage regulator (Figure 2). The basic structure of AGIs shown in Figure 1 does not
However, the development of novel control schemes illustrate the many different permutations of individual
and the analysis of system stability are fraught with many components that occur within the 200 AGIs that exist in
difficulties if attempted directly upon a real plant. Among practice.
the problems that may be encountered are excessive cost To this end the system was broken down into individual
due to plant inactivity while the identification tests neces- blocks which were then modelled and included within a
sary to tune the control schemes and/or study the plant’s fixed framework for analysis. The potential benefits of
behaviour are carried out. This is especially true when such a study are
attempting to derive transfer function relationships from
I. Individual analysis of the elements that comprise the
dynamically slow processes such as the gas transmission
total system would allow considerable simplification
system. Further difficulties arise in keeping the system
of any subsequent stability/controller design study.
variables constant while attempting to assess the effect of
II. Information gathered from identification tests upon
a specific test on the overall system behaviour. In particu-
sections of the real system could be incorporated for
lar, the gas systems considered here will “float” upon the
evaluation or updating of the model.
prevailing pressures, whose values differ continuously with
III. The effects upon the system of new and enhanced
demand. Finally the operation of all tests are limited
equipment can be investigated prior to its application.
within a specific range determined by the plant’s safety
system, which when triggered will significantly change the
system characteristics and invalidate the exercise (for ex- 2.1 Regulators
ample, the high- and low-pressure overrides in an AGI).
Because of these disadvantages, it is often desirable to The regulating control valves are arguably the most im-
develop alternative schemes initially using a model of the portant item of hardware within the AGI station. As a
process. consequence, various sizes and types of regulator are
The word ‘modelling,’ in an engineering context, has employed within the National Transmission System/Pres-
two principal meanings. The first (see Section 2) is mathe- sure Reduction stations, depending upon the policies and
matical modelling, in which the system to be studied is politics at the time of construction of the particular au-
represented by a set of equations developed by applying thority, as well as the predicted loading requirements of
the appropriate physical (and chemical) laws to each com- the future.
ponent. Solution of this set of equations to various sorts of Because the fluid is a gas, inertial effects in the flowing
stimuli will then represent the behaviour of the system. medium can be neglected and the valves characterised by
These equations, together with the computational means static and dynamic testing.

Figure 2. Illustration of hunting instability in an AGI station. Results from a field test upon an axial flow regulator.

742 Appt. Math. Modelling, 1996, Vol. 20, October

 Modelling of a two-stage high-pressure gas reduction station: I. Fletcher et al.

2.2 Steady-state jlow characteristics For liquids, C, is the standard liquid sizing coefficient,
which can be determined directly from water tests. How-
Various formulae have been proposed4mh for predicting ever, to determine this coefficient for gases, we must look
the steady-state flow rate of gases through the many at the results of air tests when incompressible and com-
different types and sizes of commercially available regula- pressible flows are equal. That is when
tor. Most are developed from Bernoulli’s or the continuity
equation and differ primarily in the pressure term that is
incorporated to describe gas density effects, although some PI - p2
~ < 0.02
are empirical relationships derived from air testing results.’ P,
Most of these formulae are intended for use over a
restricted range of pressure drop conditions, and a more Since under these conditions sin(x) =x, equation (1) can
general expression to cover the flow of a compressible be reduced to
medium under isentropic conditions must take account of
the transition in a flow regime which occurs at a critical
ratio of inlet and outlet pressures where the fluid passing
through a controlling aperture reaches sonic velocities.
Theoretical treatments of this phenomenon lead to com-
plex expressions which are computationally cumbersome.x
By introducing sufficient redundancy into the formulation, whereas the coefficient C, is dependent upon the critical
Buresh and Schuder” have proposed a method that is flow or choked flow (Q,> through the regulator at a
applicable to virtually all types and sizes of regulator particular stem position, that is when sin(x) + 1. Hence
existing today. Utilisation of the universal gas sizing equa-
tion within the proposed valve model would permit the
use of a fixed structure subroutine, whose predicted out-
put flow would be dependent upon the surrounding sys-
tem pressure, P, and Pz (psi absolute); temperature T(“R);
and the various constants necessary to describe the partic- Therefore the size of the valve is expressed in terms of the
ular regulator. The basic equation is given by maximum values of C, and Cg (for 100% opening) and
the valve type via the shape of the C, and C, characteris-
tics against stem position, x (see Figure 3). This behaviour
Q= F .C,C;P,.sin[8]sft”/hr is then introduced into the mathematical model using
polynomial fits.
where

2.3 Actuator behaliour

(1)
Although many different types of actuator are used within
the gas industry, the most prevalent device is the pneu-
within the limits 0 and f matic actuator. Ideally, the actuator stem travel is propor-
tional to the pressure in the actuator chamber. However,

where P, and Pz are the upstream and downstream valve

pressures, respectively, and C, and C, are coefficients
which are dependent upon the valve’s stem position, the
first indicating the basic flow capacity of the valve for the
critical drop, whereas the latter measures the extent of the
pressure recovery of the valve. The inclusion of constants Pwu*g

C, and G allows for the various types of gases that can be 01

passed (see Table 1). USdM

Fbw

Table 1. Gas constants for the universal gas sizing equation

Correction factor for

Specific density specific heat variation
Gas (G) (C,)

Air 1 .oo 1 .oo

Hydrogen 0.07 1 .oo
Methane 0.55 0.98
Natural gas 0.60 0.98 0 10 10 30 40 50 a0 70 00 00 1M
(Bacton) PsmMwm Vah Slam Tmvl
Nitrogen 0.97 1 .oo
Figure 3. Normalised C, and C, characteristics for a 10”
Oxygen 1.10 1 .oo
Fisher V25 regulator.

Appl. Math. Modelling, 1996, Vol. 20, October 743

Modelling of a two-stage high-pressure gas reduction station: I. Fletcher et al.

in practice, its behaviour will also be affected by friction One major advantage of this approach is that the resulting
effects, especially at the gland (pressure seal) and by fluid mode1 is well suited to the application of standard linear
forces acting on the internal components.‘0 To reduce control design techniques.
these adverse effects, feedback is commonly used in the The nonlinear dynamics of each pipeline element is to
form of a valve positioning network. be modelled using a resistance/capacitance equivalent
The speed of response of the valve/converter arrange- network, since in the presence of pipeline wall frictional
ment is limited by the rate at which the actuator chamber effects and pressure drops inertial effects may be consid-
can be charged and discharged. To make these rates ered negligible.”
similar booster relays are often used to permit greater
flows into the chamber by working from larger pressures,
as shown in Figure 4. This allowing the use of first/sec- 2.5 Pipeline restrictions
ond-order transfer functions to describe the almost linear
The pressure/flow relationship of compressible gas prop-
behaviour of the actuation system. agating down a pipeline is known to exhibit a high degree
of nonlinearity. Even in the steady state, numerous rela-
2.4 Pipeline modelling tionships have been proposed which evaluate the pressure
drop corresponding to a given flow magnitude and direc-
The dynamics of the system downstream of the second tion.2’ Any such eq uation which is to be employed to
regulator plays a major role in determining the overall estimate the discretized pipeline’s pressure profile should
stability of the network. It is therefore necessary to ensure incorporate the following:
that the model used to represent this part of the process
1. wall frictional forces, which are one of the major con-
in the simulation will reproduce the salient features ade-
tributions towards the pressure drop (P, - P,> within a
quately”-‘” without incurring an inordinately high cost in
pipeline;‘”
computing time.
2. the pressure head that occurs due to the geography of
Pipelines constitute distributed parameter systems, and
the pipeline.
a rigorous analysis involves partial differential equations
which must be solved under two-point boundary condi- One such equation that satisfies the above criteria, and
tions.‘4m” This is, in computer terms, a notoriously time- whose accuracy has been proved in practice” over medium
consuming procedure. We shall, therefore, explore the and high pressure ranges, is the genera1 panhandle equa-
commonly applied approach of converting the partial dif- tion:”
ferential equations into a set of simultaneous ordinary
equations by finite ‘differencing.‘20 In effect, this means
discretizing the pipeline into a number of sections, n.‘4X’y,21 .DS (5)
It is known that putting n = 1, that is, treating the pipeline .f I'
as a single fixed volume, oversimplifies the situation to a
misleading extent. l4 The first task is, therefore, to find the where K’ is a constant,
minimum value for n which will yield acceptable results.
0.031055.G.(h,-h,).PAVG?
E, = (6)
Z AVG . *A”G

P,, T, are the base pressure and temperature, respec-

tively; PAvo, zAvo, *AVo are the average pressure, com-
pressibility factor, and temperature of the gas; and L, D
are the pipeline dimensions of length and diameter, re-
spectively.
The correction for elevation changes (E,) is based
upon average gas density, and the pipeline wall friction
factor (f) is determined from a Moody chart using the
following equation, which evaluates the system Reynold’s
number for natural gas with a specific density (G) of 0.6
(see Table I):

Re = 8880.2
[
%
1 (7)

The friction factor can also be evaluated when under fully

turbulent flow conditions via

Figure 4. Dynamic response of a Fisher pneumatic actuator 1

fed by a Fairchild 5200 with booster and positioner. Perfor- - = [4.1og(3.7 * Relative roughness)? (8)
mance is contrasted with a linear first-order model. f

744 Appl. Math. Modelling, 1996, Vol. 20, October

 Modelling of a two-stage high-pressure gas reduction station: I. Fletcher et al.

c48STellumm lae.iaum
The above equations provide the basis for evaluating the 4cm

pipeline’s steady-state pressure profile at some specific 3%.

flow level, under the following assumptions:” Tmv 300.

w 2
I. The properties of the gas do not vary significantly zm-
over a cross section of the pipe. Irn-
II. The area of a cross section of pipe is constant. im-
III. The temperature remains constant’s (Figure 5, top) So-
T = 520”R). 0
0 1D 20 30 40 so 60
IV. The pipe has a relative roughness of 12,500 (f = vo)mmc Flw Ral#(mwuc)
0.002872). @ Qa* LknHyDa4atim
V. The pipe is horizontal.
so-
MY -
Under the above assumptions equation (5) reduces to the
WW 40
following equation, which is used to evaluate pipeline
Jo-
initial conditions for a given flow value:
20.

10.
L.Q2
Pf-P,2= (9) 0
10 20 30 40 5a 60
0
735.5812D5 v-s FlowRd. (mxwc)

Figure 5. Temperature and density variations of a typical pres-

All quantities are in SI units. sure reduction station.

where
2.6 Pipeline capacities
P,.RT P,
To complete the pipeline model, a method of evaluating K* = pb’K = - = 7 = constant
V
the pressures at the nodes is required. This was developed
from standard thermodynamic theory using the Ideal Gas P, being the base pressure and V the pipeline section
Law, PV= mRT. Differentiation of this equation with volume. Note that P, can also be determined as a func-
respect to time results in tion of the particular gas flowing.

dP 2.7 Number of stages required?

_ =K.k (10)
dt
To satisfy the transient accuracy of the discretized pipeline
model an infinite number of stages II are necessary.
where k is the mass flow rate, under the following However, computationally this is impossible because of
assumptions:
I. Negligible temperature changes Pip&m
sinxllumforn=1,2,4&1cI
II. Constant enthalpy (no external heat added)
III. Constant specific volume, hence constant specific
density, its reciprocal
Support for the above assumptions is provided in Figure 5,
which illustrates the variations in temperature and density
that occurred during the testing of a 4” axial flow valve.‘O
The tests were carried out for flow variations of up to 55
m3/sec (7 Msft3/hr), at valve differential pressures rang-
ing from 350 KN/m2 (50 psi) to 2400 KN/m2 (350 psi). It
is evident from these results that only small changes in
system temperature and density have occurred, providing
justification of the above assumptions.
Under the constant specific density pi, assumption,

Combining this with the Ideal Gas Law results in13

Figure 6. Evaluation of the optimum number of stages (n)
dP necessary to reflect pipeline dynamics. Time record shows the
- =K*.Q
(11) valve throughput of the Winkfield/Ripley telemetry control
dt scheme to a 20% load flow disturbance.

Appl. Math. Modelling, 1996, Vol. 20, October 745

Modeling of a two-stage high-pressure gas reduction station: I. Fletcher et al.

-1 0 1 2 5 -1 0 1 2
REM REAL

Figure 7. Frequency response of the AGI mathematical model.

Figure 8. Schematic of the AGI test rig.

the excessive amounts of computer programming re-
quired, indicating the need to determine the minimum
number of stages which would satisfy the accuracy/com- The trial simulation utilised twin Fisher V25 throttling
putational efficiency balance. ball valves, 12”M interconnecting pipework of length 14
This was investigated by using a simulation of the diameters, and 3 miles of IS”0 downstream pipeline. The
Winkfield-Ripley remote boundary pressure control pressure reduction was from a 6900 RN/m2 (100 psiG)
scheme by comparing the system responses, when the supply pressure to 3480 RN/m’ (500 psiG), with 50% of
system is experiencing a +20% load flow change, for the total drop occurring across each regulator, and a
various values of n. Figure 6 illustrates the responses nominal station throughput of 39.3 m3/sec (SMsft”/hr).
derived from the above test. From these responses it was To develop the transfer functions in terms of the Laplace
found that for four or more stages the differences be- operator s that describe the behaviour of the simulation
tween the simulations were sufficiently small as to be at the above operating points the nonlinear simulation
considered negligible. Further conformation of this result was linearised using a small-scale linearisation approach.
for the ‘optimum’ number of stages was found in a similar Figure 7 illustrates the frequency response of these lin-
study" upon a pipeline model employing continuity equa- earised transfer functions in the form of a Nyquist array.
tions for the pressure/flow relationships. Here element i,j refers to the frequency response be-
By keeping II to a minimum, without significantly de- tween the jth input and the ith output with its associated
grading the pipeline’s performance, the order of the sys- transfer function denoted by Gij(s) (see Table 2).
tem transfer function developed would also be kept to a These linearised models have been used extensively in the
minimum, ensuring that any study performed upon the development of single input-single output (SISO) and
AGI simulation would be as uncomplicated as possible. multi-input-multi-output (MIMO) control strategies to
provide robust control and improved performance.24-2”

2.8 AGI modelling

The AGI network shown in Figure 1 is capable of mod- 3. Practical evaluation using an experimental test rig
elling any of the stations that exist in practice. However,
It is well known that the analysis of a nonlinear system is
to simplify the simulation and minimise the size of the
fundamentally difficult for all but the simplest of cases.
resulting transfer function the safety systems were omitted
Features which identify the behaviour of these systems
and only the normal station operation was considered.
include limit cycles, chaos, catastrophes, switching, and
sliding. Any information that can be gathered about the
operation of such a system prior to any analysis is of great
Table 2. AGI multivariable control notation advantage.
One approach, often favoured by engineers, for obtain-
Input output
ing such supplementary information about the system is
1 Monitor’s input Interstage volume by construction of an experimental test rig, the test rig
signal (U,) pressure (P,) providing the hardware which can then be used to deter-
2 Active’s input Station inlet mine numerical values of the constants and other parame-
signal (U,) pressure (P,,) ters needed for the system identification. In addition, the

746 Appl. Math. Modelling, 1996, Vol. 20, October

 Modelling of a two-stage high-pressure gas reduction station: I. Fletcher et al.

Table 3. AGI test rig hardware

Variable Description Device

Inputs Ul Valve 1 (upstream) JLIN AT0 Platon M-valve (via

Fairchild T5200 I/P)
u2 Valve 2 (downstream) JLIN AT0 Platon M-valve (via
Fairchild T5200 I/P)
u, Load regulator AT0 Platon M-valve (via Fairchild
T5200 l/P)

outputs p, Station inlet pressure Sensym LX1 820GE (O/l 00 psiG)

PJY,) Interstate volume pressure Sensym LX1 820GB (O/l 00 psiG)
P,“(YZ) Station outlet pressure Sensym LX1 820GB (O/100 psiG)
Pd System downstream load pressure Sensym LX1 820GB (O/l 00 psiG)

dPg Station throughput Druck PTXllO (O/75 mBar via Dall

tube)
Pact, Valve 1 Actuator pressure Sensym LX1 820GB (O/l 5 psiG)
Pact2 Valve 2 actuator pressure Sensym LX1 820GB (O/l 5 psiG)
HLP/19O/FLl/l80/4K (O/100 mm)
XI Valve 1 stem displacement Penny and Giles
x2 Valve 2 stem displacement Penny and Giles
HLP/l 9OFLl/l80/4K (O/l 00 mm)

test rig provides a means of verifying the various mod- large volume cylinder which helps provide the relatively
elling exercises and allows the results of any analysis to be slow dynamics of the downstream system. Table 3 defines
evaluated using real data collected from a controllable the measurements and transducers/equipment used on
safe environment at minimum cost/time. the rig, all of which are converted electronically to O/10 V
outputs to match standard data acquisition cards. The
various measurements taken allow us to investigate regu-
3.1 AGI test rig lator performance as well as the various control configura-
tions that AGI stations in which can be operated.
The test rig (F&-e 8) uses the basic AGI structure as
illustrated in Figure I and operates from a supply pressure
of 690 KN/m* (100 psiG) and at flow rates up to 100 3.2 Test rig results
litres/min in 6 mm (l/4” ) diameter pipework. A solenoid
valve is employed as the slamshut, and a pair of Platon Initial testing commenced using the previously defined
M-valves provide monitor and active regulation into a pressure control strategy with regulator set points of 75%
and 50% of the station inlet pressure. Figure 9 shows the
results of open loop step responses collected after the
system had been established at the desired steady-state
levels with the load regulator 50% open. (Note that this
test would take significantly longer on the real system and
the data collected would often be corrupted by changes in
consumer demand.)
The AGI test rig was modelled in the form of the discrete
time MIMO transfer function matrix and follows the
previous elemental notation:

P” G,Jz) G,Jz) U,
P IN = G&) G&) ’ u,
II I

Each element of the matrix G(z) was determined using

recursive least-squares identification using the structure
(sampling time = 2 sets)

b,z-’ + b,zm2
G,j(Z) =
-0.21 I 1 -a,z-I -a,z-*
0 m 40 60 50 100 im

Figure 9. Normalised open-loop step responses of the AGI test resulting in the parameter in Table 4.
rig. Results show both the plant response and its linear model Using MATLAB the above data were used to deter-
(detailed below). mine the rig’s Nyquist array (Figure IO). Comparison of

Appl. Math. Modelling, 1996, Vol. 20, October 747

Modelling of a two-stage high-pressure gas reduction station: I. Fletcher et al.

Table 4. Test rig model parameters

Multivariable Transfer Function Element

G,,(z) G,2(~) G,,(z) G,,(z)

Parameter b, 0.1409 -0.1343 0.0013 0.009 1

value b, -0.1328 0.1276 0.0099 0.0075
al 0.9530 0.9122 0.4966 0.4957
a2 0.01 14 0.0047 0.4374 0.4277

nEl
1 -11 1 &molt12 A further indication of the modelling philosophy’s ac-
j /
j : / curacy is provided from the study of remote boundary

/ f
j i 1
0.5 .___.... i _......_. I__
__....__:
........ 0.5 _...._..: __“_‘__‘_~““._._..,..‘.“.. pressure control systems. 27.28 Here, models developed us-
IMA(3 I i j ing the same mathematical relationships employed in the

cr/
0 . . . . . . ..I . . . . . . . . . . . . .._.... i . . . . . . . . 0 _.__.___
+ . . . . . . . i. . . . . . . . .L. . . . . . . . construction of the AGI simulation have been used to
j i 1 design novel control schemes; the resulting controller
w
4.5 _.__..__ j. ..____. j__. __.__ i . . .._... 4.5 . . . . . . . . t __._.....f ..__..... ;___...._
j j
parameters are directly applicable to the actual systems.
1
I
1
:
; j :
., ., : :

Acknowledgments
The authors wish to express their gratitude to British Gas,
SERC, and the University of Sunderland for their support
of this work. Particular thanks are extended to the staff of
the Engineering Research Station at Killingworth for all
the help and guidance they provided.

REM REM Nomenclature

Figure 10. Frequency response of the AGI test rig
ctl correction factor for specific heat variation
CL’,c, valve stem position coefficients
pipeline diameter
this behaviour with that of the linearised mathematical
mode (Figure 7) for the same set point profile indicates
the similarities between the behaviour of the two model
forms. Indeed the only significant discrepancy is that of
E
K’
friction factor
specific density
constant
L pipeline length
the response between movement in the downstream con-
A4 mass flow rate
trol valve and its effect upon the interstage volume pres-
P pressure
sure, element G,,. In the mathematical model G,,(s) has
PAVG average pressure of the gas
pure derivative action, signified by zero steady-state offset
Pb base pressure
in the appropriate open-loop step responses, whereas the
test rig response displays phase advance-like behaviour. Q volumetric flow
Re Reynold’s number
T temperature
TAVG average temperature of the gas
Tb base temperature
4. Conclusion V pipeline section volume
This paper has explained the development of a detailed Z AVG average compressibility factor at the gas
mathematical model of a ‘typical’ AGI station which con-
tains sufficient structural flexibility to enable it to simu-
late any of the 200+ stations that exist in practice. References
Validation of the mathematical model was performed
using data provided by the Gas Engineering Research 1. Arden, W. J. B. and Hawkins, R. W. System software specifica-
tions and reliability analysis for the National Offtakes (AC11
Station at Killingworth, experimental tests on individual electronic control project. ERS Killingworth Internal Rept., 1983
pieces of hardware, and a specially designed experimental 2. Arden, W. J. B. GASCONTROL-a review of past, present and
test rig. Results obtained from simulations of the model future systems. North of England Gas Association, ERS Killing-
are presented, and a comparison of these results and worth Rept., 1988
3. Fletcher, I. Modelling and control of a gas pressure reduction
those obtained from the AGI test rig shows that the system. Ph.D. Thesis, CNAA, Sunderland Polytechnic, 1989
simulations are accurate for the set of test signals consid- 4. Moore, R. L. The use and misuse of pressure regulators. Instrum.
ered. Technol. 1969

748 Appl. Math. Modelling, 1996, Vol. 20, October

 Modelling of a two-stage high-pressure gas reduction station: I. Fletcher et al.

5. Recommended voluntary standard formulas for sizing control for transient flow in gas pipeline systems by variational methods
valves. Fluid Controls Institute, FCI 62-1, May 1962 -a survey. Society of Petroleum Engineers of AMIE, Paper no.
6. Brockett, G. F. Correlation of valve sizing methods. ISA Confer- SPE 4005, 1972
ence, Newark, New Jersey, April 1952 19. Isermann, R. Process fault detection based on modelling and
I. Turnquist, R. 0. Comparing gas flow formulas for control valve estimation methods-survey. Automatica 20, 387-404, 1984
sizing. ISA J. June, 1961 20. Lakshminarayanan, P. A., Janakiraman, P. A., Gasendra Babu,
8. Kay. J. M. Introduction to Fluid Mechanics and Heat Transfer. M. K., Murphy, B. S. A finite differencing scheme for unsteady
Cambridge University Press, Cambridge, 1968 pipe flows. Int. .I. Mech. Sci. 21, 557-566, 1979
9. Buresh, J. F. and Schuder, C. B. Development of a Unicersal Gas 21. Cox, C. S., Arden, W. J. B. and Fletcher, I. Investigation into the
Sizing Equation for Control Values. ISA trans. 3, 322-328, 1964 remote pressure control of gas transmission pipelines. Intema-
10. Fletcher, I. Axial flow regulator-a study. Sunderland Polytech- tional ASME Conference on Modelling and Simulation, Sorrento,
nic/ERS Killingworth Internal Report EEC/IF/1/87, 1987 1986
11 Tsai, D. H. and Cassidy, E. C. Dynamic behaviour of a simple
22. Orange, K. Flow formulae for gas distribution and transmission.
pneumatic pressure regulator. Trans. ASME Basic Eng. 83,
Yorkshire Junior Gas Association, ERS Library, Report no.
253-264, 1961
P/116.3, 1981
12 Dustin, M. 0. Analog computer study of design parameter ef-
23. American Gas Association. Steady Flows in Gas Pipelines. New
fects on the stability of a direct acting gas pressure regulator.
York, 1985
NASA Technical Note TN D6267, March 1971
13 Lee, W. F. Z., Bonner, J. A. and Leonard, R. G. Dynamic 24. Fletcher, I., Cox, C. S. and Arden, W. J. B. Multivariable control
analysis and simulation of a gas regulator. 1971 Symposium on strategies applied to a gas pressure reduction station. Interna-
Flow Control. 3. System Design and Application. Paper no. 3-3-100, tional Conference on Applications of Multit>ariable System Tech-
1971 niques, University of Bradford, 1990
14 Stoves, D., Thompson, B. and Graham, P. Simulation in the 25. Fletcher, I. and Arden, W. J. B. Tandem control of a gas pressure
design of gas pressure control equipment. 1983 International Gas reduction station. International Conference Control 91, Herriot
Research Conference Watt University, 1991
IS Guy, J. J. Computation of unsteady gas flow in pipe networks. 26. Fletcher, I., French, I. G. and Hague, M. Modelling and control
lnt. Chem. Eng. Symp. Ser. no 23, London, 139-145, 1967 of a coke oven battery. International Conference on Applications of
16. Kralik, J., Stiegler, P., Vostry, 2. and Zavorka, J. Modelling the Multicariable System Techniques, University of Bradford, 1994
dynamics of flow in gas pipelines. IEEE Trans. Syst. Man. Cyber- 21. Arden, W. J. B. and Fletcher, I. Whasset/Kendal remote bound-
net. 14, 586-596, 1984 ary pressure simulation study. North West Gas Research Rept.,
17. Anderson, T. R. and Farsoe, H. F. Estimation of the pressure in 1988
a natural gas pipeline. ACI 83 Applied Control and Identification 28. Arden, W. J. B. and Fletcher, I. Mill-Hill/Cricklewood remote
Proceedings, IASTED Symposium, 1983. Vol. 1 boundary pressure simulation study. North Thames Gas Re-
18. Rachford, H. H., Jr. and DuPont, T. A fast, high accuracy model search Rept., 1988

Appl. Math. Modelling, 1996, Vol. 20, October 749