Simulation Compresseur PDF
Simulation Compresseur PDF
Simulation Compresseur PDF
Control Systems
Submitted by
Claus Hansen
Circulation: 3
Pages: 77
Appendices: 3
Enclosed: CD-ROM
I
Abstract
Centrifugal compressors used for gas compression are complex fluid-flow machines with a limited operational
range and control of the compressors is crucial for safe and efficient operation. A model has been created for
simulation of a separation and gas compression system by using HYSYS DynamicsTM. Based on the theory for
centrifugal compressors and control theory a control strategy has been applied to the model based on the
available equipment. The model has been used to investigate how the gas compression system responds to
changes in the compressor inlet flows and conditions. The simulations performed have resulted in suggestions
to modifications to the system.
Resumé
Centrifugal kompressorer er komplekse maskiner, der ofte benyttes til naturgas kompression. På grund af et
begrænset operationsinterval er det essentielt at regulere kompressorerne for at sikre en pålidelig og effektiv
drift. En model for et separations- og gaskompressionssystem er opbygget i proces simulations-programmet
HYSYS DynamicsTM. Et kompressor reguleringssystem baseret på teorien for centrifugal kompressorer og
generel reguleringsteori er implementeret i modellen. Ved brug af modellen er det undersøgt hvorledes
kompressionssystemet reagerer på ændringer i flow og sammensætning af gassen. På baggrund af
simuleringerne er der givet forslag til modifikationer af systemet.
II
Preface
This project is the Final Thesis to become Master of Science in Oil and Gas Technology.
The project has been carried out in co-operation with the Process & Loss Prevention Department of Rambøll
Oil & Gas in Esbjerg. The basis of the project is a detailed engineering project of an oil and gas production unit.
A model has been developed using the simulation tool HYSYS DynamicsTM by AspenTech. During the project
period a license for the AspenTech’s University Package for Process Modeling has been purchased which
contains a comprehensive amount of programs for process simulation and among these HYSYS DynamicsTM.
The first year license has been paid by Rambøll Oil and Gas, for the purpose of evaluating it as a replacement to
other programs currently used at Aalborg University Esbjerg.
I express my gratitude to the people at Rambøll for support and interest in my project. Especially, I would like
to thank Carsten Stegelmann for guidance and constructive criticism and Asmus D. Nielsen for support,
especially in matters where experience is the key.
I direct my graditude to my fellow student Martin Olldag Bay for help and support throughout the project
period.
For referencing and citing the American Chemical Society (ACS) style along with the Name and Year System
[Name, Year] has been used as described in The ACS Style Guide. [Dodd, 1997]
__________________________
Claus Hansen
III
Nomenclature
A Area [m2]
c Absolute gas velocity [m/s]
Cd Discharge coefficient [-]
Cg Gas flow sizing coefficient [USPGM]
Cv Fluid flow sizing coefficient [USPGM]
cp Heat capacity at constant pressure [J/(mol·K)
d Diameter [m]
e (t ) Error at sampling time t [-]
Ft LMTD correction factor [-]
g Gravitational acceleration [m/s2]
h Enthalpy [m2/ss]
k Constant for frictional pressure loss [-]
Kc Controller gain [-]
Km Pressure recovery coefficient [-]
m& Mass flow-rate [kg/s]
M Total moments of momentum [Nm]
MW Moleweight [kg/kmol]
N Rotational Speed [rpm]
nT Polytropic temperature exponent [-]
nν Polytropic volume exponent [-]
p Absolute pressure [bar]
q Heat flow [m2/s2]
R Radius [m]
Rg Universal gas constant [J/(mol·K)]
Re Reynolds’ number [-]
T Absolute temperature [K]
Td Differential time [s]
Ti Integral time [s]
T2s Isentropic discharge temperature [K]
ΔT LM Log-mean temperature difference (LMTD) [K]
u Absolute blade speed [m/s]
U Overall heat transfer coefficient [kJ/(K·s)]
ν Molar volume [m3/mole]
V& Volume flow [m3/s]
Vu Peripheral component of the absolute gas velocity [m/s]
Vn Normal component of the absolute gas velocity [m/s]
IV
V rel Gas velocity relative to the impeller blade [m/s]
W Actual work [m2/ss]
yp Polytropic head [m] or [m2/ss]
ys Isentropic head [m2/ss]
yT Isothermal head [m2/ss]
z Elevation [m]
Z Compressibility [-]
β Ratio of orifice diameter and pipe diameter [-]
ε Expansibility factor of gases [-]
κ Specific heat ratio ( C p / C v ) [-]
κT Isentropic temperature exponent [-]
ηp Polytropic efficiency [-]
ηs Isentropic efficiency [-]
ηT Isothermal efficiency [-]
ηF Friction efficiency [-]
ηL Leakage effictioency [-]
ω Angular velocity [rad/s]
φ Flow coefficient, dimensionless [-]
ρ Density [kg/m3]
τ Fluid torque [Nm]
Subscripts
1 Inner or inlet
2 Outer or outlet
d Discharge
s Suction
V
Table of Contents
1 INTRODUCTION ................................................................................................................1
1.1 OBJECTIVE .................................................................................................................................................. 1
1.2 THESIS OVERVIEW...................................................................................................................................... 1
5 SIMULATION SCENARIOS..............................................................................................46
5.1 SCENARIO A ............................................................................................................................................. 46
5.2 SCENARIO B.............................................................................................................................................. 49
5.3 SCENARIO C.............................................................................................................................................. 57
6 DISCUSSION......................................................................................................................61
6.1 MODIFICATIONS TO PRESENT SYSTEM ....................................................................................................... 61
6.2 FUTURE WORK .......................................................................................................................................... 62
7 CONCLUSION ...................................................................................................................63
REFERENCES ............................................................................................................................64
APPENDIX
A SIMULATION IN HYSYS
B HYSYS DYNAMICSTM PFD FOR THE OGPU
C ANTI-SURGE CONTROLLER TUNING
VI
Introduction
1 Introduction
Centrifugal compressors in the Oil and Gas Industry constitute a vital part of process machinery at the topside
of oil and gas exploitation sites. Centrifugal compressors are expensive equipment and a major energy consumer
of the processing equipment. Thus, potential for both capital and operational savings exist, if design and
operation is dealt with in a competent way. In addition, centrifugal compressors are very sensitive to changes in
operating conditions and have a limited operational range, and if exceeded the compressor can be damaged
beyond further use. For those reasons, it is crucial that the compressors are controlled properly.
The detailed design of the compressor internals is usually taken care of by experienced vendors based on the
design basis, but the operation of them and interaction between serial and/or parallel coupled compressors and
the equipment, to which they are connected, are not. Therefore, operability studies of the system as a whole
give valuable insight into how it should be operated and controlled prior to taking it into service. In addition,
alternatives to the design can be tested.
Advanced compressor control include a number of possible control methods such as variable speed, suction and
discharge throttling, and by-pass. The ultimate goal for compressor control is to ensure safe and economical
operation while maintaining a high degree of flexibility in the system. Reliable and energy efficient anti-surge
control, load-sharing strategies, and dynamic decoupling of interacting control strategies are common features
of compressor control systems.
1.1 Objective
This project is a study of a natural gas compression system on a designed but not yet completely built Oil and
Gas Production Unit (OGPU). The design basis for the OGPU is a wellhead composition representative of the
North Sea region. It is designed for a production of up to 80 000 bbl/d of liquid (maximum 60 000 bbl/d of oil or
maximum 60 000 bbl/d of water) and 53 MMSCFD of gas. It is meant to be generic which means, that it must be
able – perhaps with slight modifications to the design – to produce from a variety of oil and gas fields at rates at
or below the above specified.
A model of the OGPU processing equipment is developed using HYSYS DynamicsTM, which is a general purpose
process simulation tool. The model includes main process equipment such as separators, pumps, and
compressors as well as piping and valves. Details such as nozzle and vessel elevations, compressor performance
curves and all vendor information readily available is applied to the model. PID controllers are added as
intended in the design and it is shown how they can be tuned using generally acknowledged methods and rules
of thumb and the compressor control strategy is integrated. Even though, the gas compression system is only a
part of the total processing system, focus is given to this part because of its sensitivity to changes in the process
conditions.
The primary objective is to make the model as realistic as possible especially in regards to the control of the
compressors. The model is then used to test the system under various process conditions. Based on these tests
potential problems with the design are identified and modifications that can improve operability are suggested.
A secondary objective is to give an example of the extensiveness in the use of HYSYS DynamicsTM as a process
simulation tool. Aalborg University Esbjerg is currently evaluating whether the AspenTech University Package
for Process Modeling should replace the currently used programs on the master level educations. Hence, this
project can be used as part of the evaluation.
1
Introduction
and the link between them established. Beside the process oriented analysis of the compression process,
strategies for advanced compressor control are described in order to set up an appropriate compressor control
strategy.
The present report contains a section on the theory behind centrifugal compressors. This is important in order
to understand the behavior and performance of the machines. Subsequently, the theory of compressor control is
described to make it possible to create a proper control system for the compressors. Hereafter the model created
in HYSYS DynamicsTM is described with focus on the specific choices made for the general setup of the model.
Finally, results of the various scenarios are presented and suggestions to modifications to the present design are
given.
2
Centrifugal Compressor Theory
Positive displacement compressors in essence work by entrapping a volume of gas and subsequently reducing
this volume which in turn increases the pressure. Positive displacement compressors will not be covered
further in this report.
Dynamic compressors generally work by imparting movement to the gas; i.e. kinetic energy is transferred from
the machines internals to the gas. By subsequent reduction of this velocity the kinetic energy is converted into
potential energy – pressure.
The two main types of dynamic compressors are:
• Axial Compressors
• Centrifugal Compressors
As the name implies axial compressors impart movement to the gas in the axial direction. This is done by a
series of rotors similar to those seen at the air intake in the front of jet-engines. Each rotor is followed by a
stator where the kinetic energy, imparted to the gas by the rotor, is converted into pressure. Axial compressors
will not be covered further in this report.
Centrifugal compressors, on the other hand, work by imparting movement to the gas in radial direction by an
impeller. This outward velocity is then converted into pressure in a diffuser. The main components of the
compressor will be described in more detail later in this section.
In this section the following subjects will be covered:
3
Centrifugal Compressor Theory
Diffuser channel
Casing
Plenum inlet
Balance Piston
Impeller
Shaft
Collector volute
Labyrinth Seals
Return inlet
The compressor in figure 2.1 has five stages. Each compression stage consists of an impeller, a diffuser channel,
a return bend, and a return inlet (or collector volute, for the last stage). Following the gas through the
compressor it enters through the inlet nozzle and is distributed around the shaft in the plenum inlet. The gas is
then led into the impellers where it is accelerated up to high velocities (often up to 2-300 m/s). In the diffuser
channel the gas is decelerated whereby the kinetic energy is converted into potential energy – i.e. pressure. The
return bend and the return inlet lead the gas on to the next stage in the compressor. After the last impeller stage
the gas is collected in the collecter volute and via the outlet nozzle it is sent for further processing.
Often compressors are connected in series and each compressor denoted as a stage. However, to differentiate
between the internal and external number of compression steps, the internals will be referred to as stages and
the entire compressor (as in figure 2.1) will be referred to as a section.
Impellers
The impellers are as mentioned before the component in which the rotational energy of the compressor is
transferred to the gas. The impeller consists of a disk with blades mounted to it and usually a cover disk is
4
Centrifugal Compressor Theory
welded or brazed onto the impellers. In the compressor the disk is mounted onto the shaft and the gas is fed
through an opening near the shaft.
The factor determining the appropriate design of the impellers are the dimensionless flow coefficient defined as
V&s
ϕ= (2.1)
4 d 2 ⋅u 2
π 2
The blade curvature and the influence on the overall performance will be covered further in section 2.3.1.
Labyrinth seals
For each succeeding stage the pressure increases and therefore it is necessary to seal the interface between the
dynamic and the stationary components. The simplest of these inter-stage seals is the labyrinth seal that work
by creating turbulence in the cavities and thereby restricting flow from the high pressure side to the low
pressure side. Seals of this type are favorable because there is no contact between the stationary and the moving
parts; hence there is no mechanical friction or wear. The principle is shown in figure 2.3.
5
Centrifugal Compressor Theory
Labyrinth seal
Shaft
Leakage Gas Flow
The seal in figure 2.3 is a knife-edge seal. The number of sealing tips in the seal depends on the application.
More sealing tips gives lower leakage flow and typically interstage seals have 5-7 sealing tips, whereas the
balance piston seal have 20-30 sealing tips. [Lüdtke, 2004]
Balance Piston
Throughout the compressor section the pressure increases for each stage. Hence a net thrust along the shaft of
the compressor exists in the direction of the compressor inlet. To outbalance the majority of this thrust a
balance piston is placed after the last stage. The area behind the balance piston is connected to the inlet of the
compressor with a line creating a pressure difference over the balance piston in the opposite direction. The
thrust not balanced out by the balance piston is absorbed by thrust bearings.
2.2.1 Thermodynamics
The general energy balance for compressors may be written in differential form as
c2
dy + dq = dh + d + g ⋅ dz (2.2)
2
Where y : Specific mass referenced compressor work input
q: Heat flow into the compressor through walls
h: Enthalpy of gas
c: Absolute gas velocity
g: Gravitational acceleration
z: Elevation
6
Centrifugal Compressor Theory
Thus, the sum of the specific mass referenced work input and the heat flow to the compressor is equal to the
sum of the enthalpy change, the kinetic energy change, and the static head difference.
Usually the velocity term and the static head contributions are regarded negligible and the compressor is
considered adiabatically isolated from the environment. Hence equation 2.2 reduces to
dy = dh (2.3)
The change in enthalpy of the gas is
dp
dh = (2.4)
ρ
Where ρ : Density of the gas
p: Pressure of the gas
The specific compressor mass referenced work input is then
p2
dp
y = ∫
p ρ
(2.5)
1
The actual work is found by dividing the mass referenced work input by an efficiency
y
W = (2.6)
η
Where W: Actual work applied to the compressor
The integral in 2.5 can be solved in different ways equivalent to different compression paths. Usually, one of
three different reference processes for compression is used to solve compressor performance. These are
7
Centrifugal Compressor Theory
κ −1
Z 1 ⋅ R g ⋅T1 ⎡ ⎤
κ ⎛ p2 ⎞ κ
ys = ⋅ ⋅ ⎜ ⎟ − 1⎥
⎢ (2.9)
MW κ − 1 ⎢⎝ p1 ⎠ ⎥
⎢⎣ ⎦⎥
Where Z : Compressibility
Rg : Gas constant
T : Absolute temperature
MW : Moleweight of the gas
p ⎛ p ⎞
p ⋅ν = const . ⇒ = const . ⇒ ρ = ρ1 ⋅ ⎜ ⎟ (2.10)
ρ ⎝ p1 ⎠
This simple relation gives a much simpler integration and the isothermal head is
R ⋅T1 ⋅ Z 1 ⎛p ⎞
yT = ⋅ ln ⎜ 2 ⎟ (2.11)
MW ⎝ p1 ⎠
Where y T : Isothermal head
8
Centrifugal Compressor Theory
is the truth for most compressors within a margin of 1 % of and is as such a reasonable approximation. [Lüdtke,
2004]
It should be noted that eventhough the polytropic reference process is adiabatic it is not reversible. This is due
to a friction caused entropy increase in the process which is irreversible.
The reference processes are all related to the actual work applied to the compressor by
yp y s yT
W = = = (2.14)
ηp ηs ηT
2.2.2 Aerodynamics
The aerodynamic explanation to the enthalpy change experienced by the fluid in the impeller section is best
described by derivation of the Euler Turbomachinery Equation; hence this is done in the following. The
derivation is based on the assumption that the single compressor stage can be regarded as adiabatically isolated
from the environment.
Figure 2.4 is a cross section of the flow path from the return inlet through the impeller and into the diffuser.
Vn
Diffuser
2 R2
Impeller
Return inlet Vn
R1 1
Shaft
w
Setting up a control volume over the impeller from inlet (1) to outlet (2), the total moments of momentum
entering and leaving the control volume is given by
9
Centrifugal Compressor Theory
M 1 = ∫ ρ1 ⋅V n ⋅Vu ⋅ R1 ⋅ dA1
1 1
(2.15)
M 2 = − ∫ ρ 2 ⋅V n ⋅Vu ⋅ R2 ⋅ dA2
2 2
τ ⋅ω
= Vu ⋅ R1 ⋅ ω −Vu ⋅ R2 ⋅ ω
m& 1 2
(2.19)
− y = Vu ⋅ u1 −Vu ⋅ u 2
1 2
y = Vu ⋅ u 2 −Vu ⋅ u1
2 1 (2.20)
Thus, it has been shown how the aerodynamics is linked to the thermodynamics. In section 2.4 it will be shown
how the performance can be calculated from polytropic head and efficiency.
10
Centrifugal Compressor Theory
• Forward leaning
• Radial
• Backward leaning
The performance of these three types of impellers can best be understood by looking at the aerodynamics at the
impeller tips in terms of velocity vectors. In figure 2.5 the three situations are depicted schematically with the
resulting velocity vectors.
c2 Vrel c2 Vrel=Vn2
Vn2 c2 Vn 2 Vrel
u2 R2 u2
Vu2 u2 =Vu2
Vu2
Rotation
R1
The velocity vectors in figure 2.5 are depicted at radius R2 (i.e. the outlet of the impeller). Similar vectors could
be depicted at radius R1 (i.e. the inlet of the impeller) only these would be different in magnitude. The normal
velocity,V n2 , the peripheral velocity, Vu2 , and the absolute blade tip speed, u 2 , were all introduced in section
2.2.2. The absolute gas velocity can be defined by
c 2 = Vu +V n = u 2 +V rel
2 2 (2.21)
Where c 2 : The absolute gas velocity
V rel : The gas velocity relative to the impeller blade
Now consider a situation where the peripheral component of the inlet velocity, Vu1 , is zero, called zero the
inlet whirl. Then the aerodynamic expression for the mass referenced specific head in equation 2.20 reduces to
y ideal = Vu ⋅ u 2
2 (2.22)
The situation is of course not possible as it would require that the gas was fed at the centerpoint of the impeller
which is impossible. However, the ideal head serve to explain the basic shape of the head curves for
compressors with the three different impeller configurations listed previously.
Since the velocity relative to the impeller blades, V rel , is proportional to flow it can be stated that for radial
bladed impellers u 2 is equal to Vu2 no matter the flow and therefore the ideal theoretical head curve is flat. For
forward leaning impeller blades Vu2 increases as the flow and thereby V rel increases. Hence the ideal
theoretical compressor curve for compressors with forward leaning impeller blades is one with a positive slope.
For compressors with backward leaning impellers Vu2 decreases with flow and therefore the ideal theoretical
compressor curve has a negative slope.
The ideal and typical performance characteristics for the three impeller types are shown in figure 2.5.
11
Centrifugal Compressor Theory
Figure 2.6. Typical head curves (solid line) for the compressors with forward leaning, radial, and
backward leaning impellers. The dashed lines are the ideal theoretical head curves for the zero inlet
whirl situation.
The vertical distance between the ideal theoretical head curves and the actual performance curves describe the
losses due to friction, leakage, and incidence.
Usually backward leaning impellers are used because they generally give the best efficiency in the largest flow
interval; i.e. the actual head curve is closest to the ideal head curve in this type of impeller compared to radial
and forward leaning impellers. Forward leaning impellers are often used when the requirement of head is
dominant because this impeller type can deliver a given minimum head in a larger flow interval than the other
two impeller types.
Line connecting
Recirculating area behind
Flow balance piston
to gas inlet
12
Centrifugal Compressor Theory
As an alternative to the in-line impeller arrangement a back-to-back impeller arrangement can be utilized. The
idea is to place two sections in the same casing with the outlets adjacent to each other and with an intercooler
between the sections. This arrangement is depicted schematically in figure 2.8.
Intercooler
Recirculating flow
Section 1 Section 2
Figure 2.8. Back-to-back compressor arrangement.
The advantage of back-to-back configuration to in-line impeller configuration is that the balance piston is
replaced by a rudimentary labyrinth seal between the outlet stages of each section. Thus, the leakage flow is
only re-circulated through the latter section of the back-to-back compressor. In result the compressor power
loss due to leakage is reduced to approximately one third of the power loss in the in-line impeller configuration.
Another advantage is that the greatest part of the impeller thrust is out-balanced by the impellers themselves.
[Lüdtke, 2004]
13
Centrifugal Compressor Theory
20000 78
19000 Polytropic 77
18000 Efficiency 76
17000 75
Efficiency [%]
Head [m]
16000 74
Polytropic Head
15000 73
14000 72
13000 71
12000 70
11000 69
10000 68
2400 2600 2800 3000 3200 3400 3600 3800 4000 4200
Actual Volume Flow [m³/h]
Similar curves for the second and third stage compressors can be found in the Excel worksheet “Compressor.xls”
on the CD-ROM.
Head vs. actual volume flow curves are used because it is believed that the compressor always deliver the same
head at the same actual inlet volume flow. This is also what the Euler equation states. However, when there are
significant compressibility changes in the flow which is generally the case when the tip-speed Mach numbers
are above 0.4 and there are more than two compressor stages, the head can vary. The factors that cause the
polytropic head to vary are parameters that affect the Mach number, i.e. compressor rotational speed, suction
temperature, and gas composition. These effects are called aerodynamic stage mismatching and will be further
discussed in section 2.5.3. [Lüdtke, 2004]
Pressure
The polytropic head for compression of real gases expressed in the unit of length is given by equation 2.23.
nν −1
⎡ ⎤
Z ⋅ R ⋅T1 nν ⎢⎛ P2 ⎞
− 1⎥
nν
yp = 1 ⋅ ⎜ ⎟ (2.23)
MW ⋅ g nν − 1 ⎢⎝ P1 ⎠ ⎥
⎣⎢ ⎥⎦
14
Centrifugal Compressor Theory
⎛ n ⎞
⎜ ν ⎟
⎡ y p ⋅ MW ⋅ g nν − 1 ⎤ ⎜⎝ nν −1 ⎟⎠ (2.24)
p 2 = p1 ⋅ ⎢1 + ⋅ ⎥
⎣ Z 1 ⋅ R ⋅T1 nν ⎦
Thus the discharge pressure is a function of suction pressure, compressor head, and the composition.
Temperature
Recalling that the definition of a polytropic process is
p ⋅ν nν = const . (2.25)
Rearranging equation 2.25 by using ν = ZRT / P gives the expression in equation 2.26 [Lüdtke, 2004].
nT −1
T nT p nT −1 p nT Pm n −1 (2.26)
p nT = const → nT = = = const , where m = T
P T T T nT
nT is the polytropic temperature exponent and is different from the polytropic volume exponent. In equation
2.26 the compressibility is assumed constant.
Using the relation in 2.26 allow for calculation of the discharge temperature as shown in 2.27.
m
p1m p 2m ⎛p ⎞
= ⇔ T 2 = T1 ⋅ ⎜ 2 ⎟ (2.27)
T1 T2 ⎝ p1 ⎠
m – and thereby the polytropic temperature exponent, nT – is calculated by equation 2.28 [Lüdtke, 2004].
Z1 ⋅ R ⎛ 1 ⎞ κ −1
m= ⋅ ⎜ − 1⎟ + T (2.28)
⎜
c p ⎝ ηp ⎟ κT
⎠
Where κT : Isentropic temperature exponent defined at constant entropy ( η p = 1 )
cp : Heat capacity at constant pressure
The isentropic discharge temperature, T 2 s , is given – similarly to 2.27 – by equation 2.29.
κT −1
⎛p ⎞ κT
T 2s = T1 ⋅ ⎜ 2 ⎟ (2.29)
⎝ p1 ⎠
Rearranging equation 2.29 gives
⎛T ⎞
log ⎜ 2s ⎟
κT − 1 ⎝ T1 ⎠
= (2.30)
κT ⎛p ⎞
log ⎜ 2 ⎟
⎝ p1 ⎠
Combining equations 2.27, 2.28 and 2.30 gives
⎛T ⎞
log ⎜ 2 s ⎟
Z ⋅R ⎛ 1 ⎞
m
⎛p ⎞ ⎝ T1 ⎠
T 2 = T1 ⋅ ⎜ 2 ⎟ , where m = ⋅ ⎜ − 1⎟ − (2.31)
⎜
c p ⎝ ηp ⎟
⎝ p1 ⎠ ⎠ log ⎛ p 2 ⎞
⎜ ⎟
⎝ p1 ⎠
15
Centrifugal Compressor Theory
Volume
The actual volume flow of the discharge stream is calculated by
ρ1
V&2 = V&1 ⋅ (2.32)
ρ2
16
Centrifugal Compressor Theory
&
Set: p1 , T1 , Composition, and m
Guess: nν and nT
Calculate P2
Calculate m
Calculate V&2
No Yes
Is D < 10−12 ? STOP!
A program has been written in the Visual Basic application for Excel using the calculation scheme in figure 2.10
to calculate compressor performance with starting point in compressor curves similar to that of figure 2.9 (cf.
“Compressor.xls” on CD-ROM).
17
Centrifugal Compressor Theory
50 170
165
45
160
155
40
150
145
35
140
30 135
130
25 125
24000 29000 34000 39000 24000 29000 34000 39000
Figure 2.11. Discharge pressure and discharge temperature as a function of mass flow at different
moleweights (P1=10.3 bara, T1 = 40 °C)
T1 = 30 T1 = 35 T1 = 40 T1 = 45 T1 = 50 T1 = 30 T1 = 35 T1 = 40 T1 = 45 T1 = 50
50
Discharge Temperature [°C]
Discharge Pressure [bara]
169.5
45 164.5
159.5
40
154.5
149.5
35
144.5
139.5
30
134.5
25 129.5
24000 29000 34000 39000 24000 29000 34000 39000
Figure 2.12. Discharge pressure and discharge temperature as a function of mass flow at different
suction temperatures (MW=23.5 g/mole, P1=10.3 bara)
P1 = 9.3 P1 = 9.8 P1 = 10.3 P1 = 10.8 P1 = 11.3 P1 = 9.3 P1 = 9.8 P1 = 10.3 P1 = 10.8 P1 = 11.3
50
Discharge Temperature [°C]
Discharge Pressure [bara]
45 162.2
40
152.2
35
142.2
30
132.2
25
122.2
20
15 112.2
24000 29000 34000 39000 24000 29000 34000 39000
Figure 2.13. Discharge pressure and discharge temperature as a function of mass flow at different
suction pressures (MW=23.5 g/mole, T1=40 °C)
As seen on the graphs in figure 2.11 through 2.13 relatively small deviations in the inlet conditions have great
impact on the discharge pressures as well as temperatures.
18
Centrifugal Compressor Theory
• Surge
• Stonewall
Surge is the lower boundary of a stable flow and stonewall is the upper limit of the compressor flow. In addition
to this the effect of stage mismatching should be considered when altering the inlet conditions.
In the following these three phenomena will be explained.
2.5.1 Surge
Surge can happen on two different levels. Stage surge is surge in a single stage of the compressor whereas
system surge is when the entire system goes into surge. Although the phenomena are similar with regard to
causes the implications are somewhat different.
Stage surge
In figure 2.14 the flow through the impeller is depicted under normal operation and at stage surge.
Vgas Vrel
Vgas Vrel
a
U2 Propagation
U2 of Stage Surge
Impeller blade
Flow separation
Rotation
Figure 2.14. Flow patterns in an impeller at normal operation and at stage surge.
a: flow angle, i: incidence angle, Vrel: gas velocity relative to impeller blade, u2: impeller blade tip
speed, Vgas: velocity of gas.
As depicted in figure 2.14(a) the gas flows through the ducts between the impeller blades under normal
operation. However, when flow is reduced – or rotating velocity is increased at constant flow – the flow angle
is decreased. This means that each gas molecule must travel a longer distance and more of the flow momentum
is dissipated by friction at the impeller blades. At the same time the incidence angle at the impeller inlet is
increased which causes flow separation at the low pressure side of the impellers as depicted in figure 2.14(b).
The flow separation tends to continuously shift round the diffuser from one impeller blade to the next in the
opposite direction of the rotation. [Gresh, 2001]
19
Centrifugal Compressor Theory
The flow separation due to poor incidence angle leads to flow reversal and the stage will exhibit flow and
pressure fluctuations around a mean value. These pressure and flow fluctuations are known as stage surge. The
stage surge condition can be present both at the impellers and in the stationary flow channels, but the effect is
similar.
Stage surge occurs at low flow just before system surge occurs, but if operating conditions remain constant it is a
stable compressor operating condition with a net forward flow. However, for the component experiencing this
flow condition it is a dynamic instability, which can cause damage to the components under influence. [Lüdtke,
2004]
System Surge
As mentioned above system surge is a condition that occurs if flow is lowered even more in a condition of stage
surge. However, system surge can also occur without a preliminary stage surge.
The aerodynamic explanation to the condition is similar to that explained for stage surge, but the implications
of a system surge are much more severe as all components in the compressor can be damaged beyond the point
where operation of the compressor is possible. In system surge it is not a local flow reversal at an impeller or in
the flow channels, but flow reversal through the entire compressor.
System surge is best described by extending the operating curve into the second quadrant as depicted
schematically in figure 2.15.
P
D' D
B' B
A' A
C' C
F E
E'
Quadrant II Quadrant I
Volume Flow
Quadrant III Quadrant IV V&
Figure 2.15. Performance curve extended into the second quadrant. [Lüdtke, 2004]
Imagining a system with a pressurized gas reservoir at the outlet of the compressor the following reasoning for
operating points A, B, and C can be made with reference to the curve in figure 2.15:
1. If the operation point shifts from A to B the reservoir momentarily deliver the difference V& −V& A B
which lower the pressure of the reservoir and thereby restores the former operating point, A.
2. If the operation point shifts from A to C the reservoir momentarily stores the difference V& −V&
C A
which increase the pressure of the reservoir and thereby restores the former operating point, A.
Thus, condition A is a stable operating point and in this part of the performance curve the system is to a large
extent self-controlled. Similar reasoning for A’, B’, and C’ give different results:
20
Centrifugal Compressor Theory
3. If the operation point shifts from A’ to B’ the reservoir momentarily stores the difference V&B ' −V&A '
which increases the pressure of the reservoir and thereby moves the operating point further away from
A’. On the path from A’ to D the compressor continuously pumps more flow into the reservoir, but as
D is reached and the pressure still increases due to the increase in flow, the only possible way is for the
compressor instantaneously to shift to operating point D’. This reverses the flow and the reservoir
begins to unload and the operating point moves towards E’. Further pressure reduction is only possible
if the operating point shifts to E, and from here the reservoir begins to fill increasing the pressure, and
thus moving the operating point towards D. From here the cycle DD’EE’D begins once again.
4. Similar reasoning can be made if the operating point shifts from A’ to C’. The result is a flow cycle
E’EDD’E’.
Therefore, condition A’ is an unstable operating point. [Lüdtke, 2004]
The consequence of this is that stable operation can only be achieved of the flow is greater than that of point D.
2.5.2 Stonewall
The other limit to normal operation of a compressor is called stonewall. The phenomenon is characterized by
the sonic gas velocity at the impeller blade inlet. It occurs when the system resistance decreases and flow
thereby increases.
For a single stage compressor stonewall occurs when the head becomes zero, but for multistage compressors the
situation is different. As the inlet flow to the compressor increases each subsequent stage sees a disproportional
flow increase. That is, the relative flow increase becomes larger for each subsequent stage. Therefore, as flow
increases, the first stage in which zero head occur, is the last stage of the compressor. Thus, the compressor flow
limit – i.e. stonewall – is determined by the last compression stage. In figure 2.16, the situation is depicted
schematically for a five stage compressor.
Stonewall
Stage curves 1
2
3
4
5
Volume Flow
Figure 2.16. Individual stage head curves and total compressor curve for a five
stage compressor. The total head curve is the sum of heads of the individual stages.
A multi-stage compressor subjected to an ever decreasing system resistance can in certain cases operate stably
beyond stonewall. In this case the initial stages produce head and the later stages consume the head. However,
especially thrust bearings in the compressor must be sized accordingly, because thrust variations throughout the
21
Centrifugal Compressor Theory
stages of the compressor will be inevitable. Although, stable operation can be achieved it should be avoided
because it is a situation where much of the enthalpy created in the first stages are transferred to unavailable
energy – heat – in the later stages. [Lüdtke, 2004]
2.6.1 Serial
In essence serial coupling of compressors are similar to the division of stages. However, one major difference
between the internal stages of a compressor and serial coupling of two or more compressors is that the gas is
usually not cooled between stages, whereas cooling is usually done between sections.
Curves for the three serial connected compressors at the OGPU and the resulting total head curve are shown in
figure 2.17.
22
Centrifugal Compressor Theory
45000
40000
Total Head
35000
30000
Head [m]
25000
20000
15000
1
10000
2
3 Sections
5000
0
0 0.2 0.4 0.6 0.8 1 1.2
Actual Volume Flow [m³/s]
Figure 2.17. Polytropic head curves for the three serial connected compressors on the OGPU and the
resulting total polytropic head curve.
As explained in section 2.5.2 each succeeding stage sees a disproportional change in flow as the inlet flow is
increased or decreased. Similar reasoning can be made for compressors connected in series, and hence surge or
stonewall is most liable to occur at later sections as the flow is changed at the inlet of the first section. Hence,
surge can be expected to occur first in the last section and last in the first section.
2.6.2 Parallel
Compressors are often coupled in parallel in order to have a large turn-down ratio. Parallel connection of
compressors mainly affects the ability of the system to adapt to different flows. However, one major implication
to the control of the compression system is that inlets and outlets are in open connection to each other and
therefore pressures are balanced. Since it is impossible to construct compressors exactly similar and because it is
inevitable that piping etc. around the compressors are different, the parallel connections calls for a system to
balance the load on the compressors. This is called load-sharing and will be treated further in section 3.4.
23
Compressor Control
3 Compressor Control
Compressors are to some extent self-controlled when operating in the stable area of the performance curve as
described in section 2.5. However, to avoid operating in unstable areas of the performance curve, and to enable
off-design operation, compressor control must be integrated into the system.
In this section the following subjects are covered
• Control methods
• General approach to compressor control
Subsequently, methods to enable essential capabilities of the overall control system will be described. These are
• Anti-surge control
• Distribution of loads – i.e. load-sharing
• Dynamic de-coupling
The reason to employ anti-surge control and load-sharing capability are evident from the descriptions in
section 2.5 and 2.6. Dynamic de-coupling must be employed to eliminate regions of instability where control
strategies that are interactive can create unstable process oscillations.
The function of these control methods is to extend the operating range of the compressor beyond the single
performance curve (as depicted in figure 2.9) and to a so-called performance map. As seen from the following
this extension can be made with variable success in regards to maintain a high efficiency on the compressor.
⎛p ⎞
V& ∝ N , y p ∝ N 2, ln ⎜ 2 ⎟∝N ,
2
(T 2 −T1 ) ∝ N 2 (3.33)
⎝ p1 ⎠
Where N : Rotational speed
These relationships can be deduced from aerodynamics by assuming an ideal gas and a single stage compressor.
Thus, for actual systems with real gases and multiple compressor stages, the accuracy of the relationships are
somewhat reduced. However, they serve as a mean to show what can be expected when reducing or increasing
speed of the compressor. A graphical representation of the fan laws are shown in figure 3.18. [Gresh, 2001]
24
Compressor Control
25000
20000
Polytropic Head [m]
110%
15000
100%
90%
10000
80%
5000
0
1500 2000 2500 3000 3500 4000 4500 5000
Actual Volume Flow [m³/hr]
Figure 3.18. Application of fan laws on the first section compressor curves of the OGPU at speeds of
80-110 % of design speed.
As stated above the fan laws are applicable to ideal gases and single stage compressors, only. The implications of
real gases and multiple stage compressors are the effect of aerodynamic stage mismatching as explained in
section 2.5.3.
The effect of stage mismatching on the head curves as shown in figure 3.18 is that the 80 and 90 % curves in
reality lie lower and farther to the left, and the 110 % line lie farther up and to the right.
Advantages to using variable speed to control compressor performance are:
• High part-load efficiencies (>95 % of design efficiency) as compressor only produce the head necessary
in part-load situations
• Possible to overload the system (by volume flow) because of over-speed
• Suitable for all compressor types
The main disadvantage is that a driver with variable speed is required. However, for most applications variable
speed control is the first choice. [Lüdtke, 2004]
25
Compressor Control
However, the low efficiency at part-load, suction throttling is widely used especially due to its simplicity and
low investment costs. [Lüdtke, 2004]
• Medium part-load efficiencies (lower than variable speed, higher than suction throttling)
• With negative prerotation of the IGV’s it is possible to overload
The use of adjustable IGV’s is mainly in situations where it is possible to fit IGV’s at more than one stage (i.e. in
integrally geared compressors).
Because by-pass is the only control method that expands the performance map to zero flow it is always an
integral part of the compressor control system.
26
Compressor Control
V&12 ⋅ MW Δp Δpo ,s
q12 ≡ ∝ o ,s ⇒ q 12 = C 1 ⋅ (3.34)
Z 1 ⋅ R g ⋅T1 p1 p1
2 ⋅ Δpo C d ⋅ ε ⋅ Ao
V& = ⋅ (3.35)
ρ 1− β 4
27
Compressor Control
0.9
0.85
Discharge Coefficient (C d)
0.8
β = 0.50
0.75 β = 0.55
β = 0.60
β = 0.65
0.7 β = 0.70
β = 0.75
0.65
0.6
0.55
1000 10000 100000 1000000 10000000
Reynolds Number
Figure 3.19. Discharge coefficient as function of the Reynolds number. L1=L’2= 0.2
Thus, in most applications relating to gas flow measurements in connection with compressors it is sufficient to
assume a discharge coefficient of 0.6.
The expansibility factor accounts for the compressibility of the gas and can be calculated using the empirical
relation
ε = 1−
( 0.41 + 0.35 ⋅ β ) ⋅ Δpo
4
(3.37)
κ ⋅ p1
Where κ : Specific heat ratio ( C p / C v )
p1 : Pressure upstream orifice
28
Compressor Control
For most gas flow measurement applications the pressure drop across the orifice is low compared to the pressure
upstream the orifice that the expansibility factor is close to unity (i.e. the second term in equation 3.37
approaches zero).
Anti-surge control is only a protective system and should as such mainly be activated when all other means
extending the performance map has been utilized.
Using the reduced flow vs. pressure ratio coordinate system for anti-surge control a common control system
takes starting point in assuming that
p2 Δp
= C 2 ⋅ o ,s + 1 (3.39)
p1 p1
Thus, it is assumed that the anti-surge control line satisfy equation 3.39.
Typical compressor performance curves and the surge control lines obeying equation 3.39 can be seen
schematically in figure 3.21.
29
Compressor Control
p2
p1
Compressor Performance
q 12
30
Compressor Control
31
Modeling of the OGPU
4.1.1 Separation
The processing equipment for separating the oil, gas, and water consists of a separation system including two
separators in series followed by an electrostatic coalescer.
The first stage separator is a three phase separator designed for a flow of 68000 bbl/d of oil, 60000 bbl/d of water
and 56.5 MMSCFD of gas. This design oil and gas flow is greater than the 60000 bbl/d and 53 MMSCFD which
is the overall design flow because of recycle from compressor suction scrubbers. The operating pressure and
temperature of the 1st stage separator is 10 barg and 50-90 °C depending on the inlet composition and
conditions. An inlet heater allow for heating of the incoming wellhead fluid.
Oil from the 1st stage separator is passed on to the 2nd stage separator which is a two phase separator designed
for a gas flow of 6.3 MMSCFD and a liquid flow of 67000 bbl/d. The operating pressure and temperature is 1
barg and 75-90 °C.
Liquid from the second stage is separated into oil and water in the electrostatic coalescer designed for a liquid
capacity of 60000 bbl/d. The purpose here is to remove the remaining water before the stabilized oil is cooled
and pumped to the storage tanks. The crude oil must be stabilized to a Reid Vapor Pressure (RVP) less than 5
psia (@100 °F).
32
Modeling of the OGPU
Figure 4.23. Process Flow Diagram of separation and gas processing equipment at the OGPU.
33
Modeling of the OGPU
Table 4.1. Design cases utilized in HYSYS steady state model. Flows defined at upstream choke
valve pressure of 15 bara.
The feedstream composition of oil and gas at standard conditions are equal regardless of the oil/water ratio.
However, the temperature and the oil/water ratio give some differences in the gas composition at the
compressors. The differences can best be seen by looking at the molewights of the gas at the compressors.
Results from the steady state model are given in table 4.2.
Table 4.2. Moleweight of the gas at the compressors from HYSYS steady
state models.
Moleweight at compressor
Case
1st section 2nd section 3rd section
Max Oil – High Temp. 23.64 22.83 22.83
Max Oil – Low Temp. 23.33 22.68 22.68
Max Water – High Temp. 23.38 22.61 22.61
Max Water – Low Temp. 22.07 21.71 21.71
34
Modeling of the OGPU
The HYSYS DynamicsTM model is based on the “Max Oil – High Temperature” case. This has been selected
because the compressor performance curves delivered by the vendor are based on a 1st section moleweight of
23.8 and 2nd/3rd section moleweight of 23.0. Thus, the “Max Oil – High Temperature” case comes closest to the
conditions under which the performance curves have been generated.
The wellhead fluid composition utilized for the HYSYS DynamicsTM model is given in table 4.3.
The pseudo components are characterized by the parameters given in table 4.4.
35
Modeling of the OGPU
Table 4.5. Boundary conditions applied to the HYSYS DynamicsTM model for
the OGPU.
As seen in table 4.5 the boundary condition of the well fluid is defined by flow, instead of pressure, which is
generally recommended in HYSYS manuals. This is done because it resembles the actual system better than
using pressure as the stream specification. In addition, it is possible to put in algorithms to vary the inlet flow
for simulation of e.g. slugs. The use of pressure or flow specifications is discussed further in appendix A section
A.3.2.
36
Modeling of the OGPU
Tests have been performed to select appropriate time step and execution rates for the HYSYS DynamicsTM
model of the OGPU. These tests are based on the conditions at the 3rd section compressor (26-KA-003A)
because here inaccuracies are assumed to be most significant. Results of suction and discharge pressure, suction
temperature and moleweight for simulation setups using time steps 0.10, 0.25, and 0.50 seconds are shown in
figure 4.24. In addition it has been attempted to reduce the composition execution rate to 2 and 5 for setups
using 0.50 second time step. The starting points of all five simulations are equal and the simulations have been
run for 10 minutes (simulation time).
105.5 252.4
0.50 sec
0.50 sec
105.0 0.25 sec 252.2 0.25 sec
Discharge Pressure (bara)
102.0 250.8
1044200 1044400 1044600 1044800 1045000 1045200 1044200 1044400 1044600 1044800 1045000 1045200
Simulation time (s) Simulation time (s)
40.4 22.84
0.50 sec
0.25 sec
22.82
40.2 0.10 sec
Suction Temperature (°C)
0.50 s / Comp. 2
0.50 s / Comp. 5
22.80
40.0
MW (g/mol)
22.78
39.8
22.76
39.6 0.50 sec
22.74 0.25 sec
0.10 sec
39.4 22.72 0.50 s / Comp. 2
0.50 s / Comp. 5
39.2 22.70
1044200 1044400 1044600 1044800 1045000 1045200 1044200 1044400 1044600 1044800 1045000 1045200
Simulation time (s) Simulation time (s)
Figure 4.24. Integrator step size and execution rate tests. The absolute values of the simulation time
is merely a measure of how long the integrator has been running without reset.
As seen on the graphs in figure 4.24 the simulations using a timestep of 0.10 s and 0.25 s give results similar to
each other. Using a timestep of 0.50 seconds, however, seem to introduce errors in the calculation on especially
suction temperature and moleweight. The increase of the composition execution rate seems to reduce the error
giving results closer to the ones obtained using the 0.10 and 0.25 second timesteps.
37
Modeling of the OGPU
The results in figure 4.24 should be compared with the Real Time Factor (RTF) which is a direct measure of the
calculation speed. The RTF is defined as
Simulation Time
RTF= (4.40)
Computation Time
Thus, the higher the RTF values the less computation time is need for a given simulation time. Hence it is
advantageous to have an RTF value as high as possible, without compromising with accuracy.
Approximate RTF’s found for the five simulations are given in table 4.7.
Table 4.7. Approximate Real Time Factors found for the five simulation setups.
Integrator setting
Step size 0.10 s 0.25 s 0.50 s 0.50 s 0.50 s
Composition execution rate 10 10 10 5 2
Real Time Factor 2 5 10 7 3.5
An integrator step size of 0.25 seconds has been selected because it from the tests performed here give results
consistent with using a lower timestep and still has an acceptable RTF value.
Table 4.8. Rules of thumb for PID controller tuning. Selected values in parentheses.
*There are no liquid pressure controllers present in the model.
The values given in table 4.8 are applied to the model for most of the controllers and it has proven to be
sufficient. However, to optimize the compressor control extra effort has been put into tuning the pressure and
flow controllers in connection with the compressors. An approach to fine tune controllers is to increase the
proportional action with integral and derivative action at a minimum. Subsequently, integral and derivate
action can be used to trim the proportional response. The optimal response can be defined as the Quarter Decay
Ratio as described in the following.
38
Modeling of the OGPU
PV
A
B
SP
Time
The final control elements (FCEs) can also be seen in figure 4.9 (PFD for the OGPU) and they are in all cases
control valves. The controllers are ordinary PID controllers receiving their control signal from the transmitters
and controlling the openings of the FCEs. The controllers and transmitters can all be found in the HYSYS
DynamicsTM model Process Flow Diagram as seen in appendix B.
Thus, the entire control system developed in the HYSYS DynamicsTM model is based solely on the possibilities
for measurement and control provided by these elements. The actual control system, however, is part of
packages delivered by external vendors and therefore the control system applied to the model might not reflect
the intentions in the final design/setup of the control system.
The control system in the HYSYS DynamicsTM model has the following features
39
Modeling of the OGPU
The setup of these control features are described in the following paragraphs.
Antisurge control
The anti-surge control system is based on the pressure ratio vs. reduced flow rate coordinate system (as
described in section 3.2.1).
Since the actual volume flow in the model is measured directly from the material stream – and not by means of
an orifice flow meter – it has been necessary to make some preliminary computations to convert the model
measurement of flow into what is really measured: the pressure drop across an orifice at suction. The
computations are performed in the surge-control spreadsheets located next to the anti-surge controllers as seen
on the PFD in appendix B.
By rearranging equation 3.35 for flow through an orifice the pressure drop across the orifice is calculated as
V& 2 ⋅ (1 − β 4 ) ρ
Δpo = ⋅ (4.41)
(C d ⋅ ε ⋅ Ao ) 2
2
The above equation can be solved by simply assuming a discharge coefficient, C d , of 0.60 and an expansibility
factor, ε , of 1.0 as described in section 3.2.2. The gas density, ρ , and the actual volume flow-rate, V& , are
continuously read from the material streams in HYSYS. Values for the factor, β , has been presumed, and are as
such not based on actual design data. Thus, the pressure drop across the orifice is continuously calculated by
readings of actual volume flow and density in the HYSYS DynamicsTM model.
The orifice pressure drop is subsequently used to calculate the Process Variable ( PV ) of the anti-surge
controller which is the reduced flow rate squared. This is done by using equation 3.34 with the constant C 1 set
arbitrarily to 106 for all compressors. Thus, the PV is calculated as
ΔPo ,s
PV = 106 ⋅ (4.42)
P1
ΔPo ,s ⎛P ⎞ 1
SP = = ⎜ 2 − 1⎟ ⋅ (4.43)
P1 ⎝ P1 ⎠ C2
The calculated values for SP and PV are both imported to a PID-controller where an Operating Variable
( OP ) is calculated by the characteristic control equation
Kc de (t )
OP (t ) = OPss + K c ⋅ e (t ) +
Ti ∫ e (t ) dt + K c ⋅Td
dt
(4.44)
40
Modeling of the OGPU
Td : Differential time
HYSYS DynamicsTM solves the controller by using a discretized version of equation 4.44. This is discussed in
more detail in Appendix A section A.3.3.
As discussed in section 3.3.1 the selection of C 2 determines the SP of the anti-surge control system.
C 2 has been selected by trial and error to enable anti-surge control at the flows given in table 4.10.
As seen in table 4.10 the selection of constants C 2 has given surge control lines in a relative distance of 8-10 %
from the actual (presumed) surge points. Exact points of surge has not been provided by the compressor vendor,
and hence the presumed surge point is the lower minimum flow reported in the performance curves delivered
by the vendor.
To have good anti-surge control, the controller must be quite aggressively tuned in order to be effective on
sudden load changes. At the same time overshoot and oscillations must be avoided.
For correct tuning of the anti-surge controllers, a series of tests have been performed. In the tests the inlet gas
flow is instantaneously reduced to 50 % of the normal gas flow. The starting point for the controller tuning is
the rules of thumb for flow controllers as given in table 4.8. Compressor inlet actual volume flows and anti-
surge valve openings versus simulation time for the 1st section compressors are seen in figure 4.26.
3400 20
3300 18
3200 16
26-KA-001A - Actual Volume Flow
Actual Volume Flow [m /h]
2900 10
2800 8
2700 6
2600 4
2500 2
2400 0
4370 4372 4374 4376 4378 4380 4382 4384 4386
Time [min]
st
Figure 4.26. Anti-surge control activation on 1 section compressors upon an inlet gas flow reduction
to 50 % of normal flow. Kc=0.5, Ti=0.15 min, and Td=0.
41
Modeling of the OGPU
As seen in figure 4.26, the anti-surge controller reacts by opening the by-pass valves as the actual volume flow
to the compressors reaches the surge control line at approximately 2700 m3/h. The difference between the flows
and responses of the A and B sections is a result of difference in piping configuration (load-sharing is
deactivated in these tests). The actual volume flow of compressor 26-KA-001B initially drops to 2600 m3/h
before stabilizing at the surge control line at 2700 m3/h. Hence the overshoot is 100 m3/h. A measure of the
relative overshoot can be defined as
= 3 = 50 % (4.45)
Anti-surge Control Flow − Actual Surge Flow 2700 mh − 2500 mh
3
Thus, a sudden reduction of the gas flow by 50 % results in an overshoot of 50 % which is unacceptable. Hence,
the anti-surge controller should be more aggressively tuned.
In figure 4.27 and 4.28 the response of the anti-surge control system for the 2nd and 3rd section compressors can
be seen.
680 30 210 30
600 10 170 10
580 5 160 5
560 0 150 0
4370 4372 4374 4376 4378 4380 4382 4384 4386 4370 4372 4374 4376 4378 4380 4382 4384 4386
Time [min] Time [min]
Figure 4.27. Anti-surge control activation on 2nd Figure 4.28. Anti-surge control activation on 3rd
section compressors upon an inlet gas flow section compressors upon an inlet gas flow
reduction to 50 % of normal flow. Kc=0.5, reduction to 50 % of normal flow. Kc=0.5,
Ti=0.15 min, and Td=0. Ti=0.15 min, and Td=0.
Calculating the relative overshoot for the 2nd and 3rd section anti-surge control systems gives 52 % and 64 %,
respectively.
In order to reduce the overshoot, while maintaining a stable non-oscillating system a series of tests have been
performed to find the optimal tuning parameters. The approach is to gradually increase the aggressiveness of
the controller until a satisfying result is obtained. A summary of the simulations performed are given in
appendix C. The selected tuning parameters are controller gains of 4 and integral time of 0.1 minutes for 1st
section anti-surge control and 0.2 minutes for 2nd and 3rd section anti-surge controllers. This gives acceptable
overshoots of 10-15 % for a sudden gas flow reduction of 50 % as seen in figures 4.29, 4.30, and 4.31.
42
Modeling of the OGPU
3400 20
26-KA-001A - Actual Volume Flow
3300 26-KA-001B - Actual Volume Flow 18
26-FV-0032 - Percentage open
3200 26-FV-0332 - Percentage open 16
Actual Volume Flow [m /h]
3100 14
3000 12
2900 10
2800 8
2700 6
2600 4
2500 2
2400 0
4398 4400 4402 4404 4406 4408 4410 4412 4414 4416
Time [min]
st
Figure 4.29. Anti-surge control activation on 1 section compressors upon an inlet gas flow reduction
to 50 % of normal flow. Kc=4, Ti=0.1 min, and Td=0. Max. overshoot 10 %.
680 45 210 50
26-KA-002A - Actual Volume Flow 26-KA-003A - Actual Volume Flow
26-KA-002B - Actual Volume Flow 205 26-KA-003B - Actual Volume Flow 45
670 40
26-FV-0056 - Percentage open 26-FV-0081 - Percentage open
26-FV-0356 - Percentage open 26-FV-0381 - Percentage open
200 40
660 35
Actual Volume Flow [m /h]
195 35
Percentage Open [%]
650 30
3
190 30
640 25
185 25
630 20
180 20
620 15
175 15
610 10
170 10
600 5
165 5
590 0
160 0
4398 4400 4402 4404 4406 4408 4410 4412 4414 4416
4398 4400 4402 4404 4406 4408 4410 4412 4414 4416
Time [min]
Time [min]
Figure 4.30. Anti-surge control activation on 2nd Figure 4.31. Anti-surge control activation on 3rd
section compressors upon an inlet gas flow section compressors upon an inlet gas flow
reduction to 50 % of normal flow. Kc=4, Ti=0.2 reduction to 50 % of normal flow. Kc=4, Ti=0.2
min, and Td=0. Max. overshoot 13 %. min, and Td=0. Max. overshoot 15 %.
As seen the overshoots are significantly reduced at the expense of some oscillation of the flow, especially on the
2nd and 3rd section. These oscillations can, however, better be accepted than a large overshoots in the case of
anti-surge control and the responses are close to having a Quarter Decay Ratio.
43
Modeling of the OGPU
open. The reason for this it is that it is not necessary to by-pass hot gas at low flow-rates due to the very limited
cooling effect at low flow-rates. Hereby it is attempted to avoid process oscillations.
With the restriction described above the hot-gas by-pass PID controllers are tuned as ordinary flow controllers
with gains of 0.5 and integral times of 0.15 minutes.
Figure 4.32. Configurations of gas compression system with suction throttling valves 26-PV-0002
and 26-PV-0302.
As is evident from figure 4.32 it is not possible to control the distribution of loads on the 2nd and 3rd stage
compressors because there is no suction throttling valve on these sections. However, as the flow between 2nd
and 3rd section is kept separate for the A and B train it would require only two suction throttling valves placed
just before the 2nd section suction scrubbers (26-VG-002A/b) to enable load-sharing at the 2nd and 3rd section.
As mentioned in section 3.4 load-sharing for similarly designed compressors should only be activated as the
flow approaches the surge control line. In the HYSYS simulation of the OGPU the load-sharing enable line of
the 1st section compressors has been established approximately 10 % above the surge control line. Thus, the
compressors are allowed to operate unrestricted until the load-sharing enable line is reached.
In order to illustrate the effect of load-sharing a case has been run where the inlet gas flow is linearly reduced
to 50 % of normal flow over a period of 30 minutes (in the simulation this is done by using the Ramp function
in a Transfer Function Block (see appendix A section A.3.3)). The actual volume flows at the compressor inlets
and the suction throttling valve openings versus time can be seen in figure 4.33. In order to emphasize the
effect of load-sharing a flow restriction has been applied to the 26-KA-001A compressor inlet in the HYSYS
model. This is the reason for the difference in flows of approximately 100 m3/h before the load share enable line
is reached. The controllers are tuned as ordinary flow controllers with a gain of 0.5 and an integral time of 0.15
minutes.
44
Modeling of the OGPU
3600 100
3400
95
3200
Actual Volume Flow [m /h]
3000
90
2800
85
2600
26-KA-001A - Actual Volume Flow
26-KA-001B - Actual Volume Flow
2400
26-PV-0002 - Percentage open
80
26-PV-0302 - Percentage open
2200
2000 75
5175 5180 5185 5190 5195 5200 5205 5210 5215 5220
Time [min]
Figure 4.33. Effect of load-sharing control on 1st section compressors. The load share is enabled at
approximately 3000 m3/h.
As seen in figure 4.33 the suction throttling valve 26-PV-0302 starts closing as the actual volume flow to
compressor 26-KA-001A reaches approximately 3000 m3/h. Full load-sharing is obtained at a compressor inlet
flow of approximately 2800 m3/h. Shortly hereafter anti-surge control is enabled and the flow is kept constant
at approximately 2700 m3/h.
Capacity Control
The suction throttling valves at the 1st section compressors can to a limited extend be also be used for capacity
control. However, they can never be allowed to close fully and in the simulation the suction throttling valves
are set to always have a minimum opening of 70 %. The valves are controlled by the suction pressures of the 1st
section compressors and in the HYSYS DynamicsTM model they are set to 12 barg.
Hence controller for the throttling valve should react on two signals. In the HYSYS DynamicsTM model the
controller is modeled by two separate controllers and the combined signal determine the opening of the valve.
However, capacity control can overwrite load-sharing control. To distinguish the controllers in the HYSYS
DynamicsTM model they are followed by a suffix of “Load” and “Pressure”, for Load-sharing control and
Pressure control, respectively.
45
Simulation Scenarios
5 Simulation Scenarios
The HYSYS DynamicsTM model for the OGPU has been used to investigate the performance of the gas
compression system under various operating conditions. Thus, three simulation scenarios have been defined.
These are
The general purpose of the simulations is to identify potential problems with the current compressor control
setup, which subsequently can be used to suggest modifications to the present layout.
5.1 Scenario A
An inlet gas flow reduction is likely to activate the anti-surge control system on both the A and B train
compressors. As described previously, the current system does not give the possibility to share loads on the 2nd
and 3rd section compressors, because a suction throttling valve (or other means for load-share control) is not
present in this system. Therefore, a gas flow reduction could result in either of the two parallel coupled
compressors to take most of the net-load, while the other recycles. If this is the case is investigated in the
following.
Action Reduce inlet gas flow to 60 % of normal flow over a 30 minute period
Compressor inlet actual volume flow
Anti-surge valve openings
Monitor
Suction Throttling valve openings (1st section)
Hot-gas By-pass valve openings (2nd and 3rd section)
5.1.2 Results
46
Simulation Scenarios
3400 100
3300 90
3200 80
26-KA-001A - Actual Volume Flow
26-KA-001B - Actual Volume Flow
Actual Volume Flow [m /h]
2900 50
2800 40
2700 30
2600 20
2500 10
2400 0
5400 5450 5500 5550 5600 5650
Time [min]
st
Figure 5.34. Results from 1 section compressors.
From figure 5.34 it is obvious that the load-sharing strategy works as the actual volume flow to the two
compressors are kept equal over the entire simulation period of 3 hours. However, the anti-surge control valve
(26-FV-0332) on the B-compressor closes after approximately one hour of anti-surge control, while the A-
compressor anti-surge control valve (26-FV-0032) opens more in that period. This does not affect the load-
sharing, but the recycle fractions on the two compressors diverge over time.
Thus, with the current configuration it can be concluded that load-sharing on the first compressor works, but
due to the placement of the suction throttling valves it is not possible to balance the recycle fractions on the
two parallel compressors. If the suction throttling valves were placed before the mixing of the incoming gas
with the recycled gas it would be possible to keep the recycle fractions on the compressors equal. The current
and the alternative configuration can be seen in figure 5.35 and 5.36, respectively.
Figure 5.35. Current configuration of suction Figure 5.36. Alternative configuration of suction
throttling valves. throttling valves.
47
Simulation Scenarios
690 100
680 90
640 50
630 40
620 30
610 20
600 10
590 0
5400 5450 5500 5550 5600 5650
Time [min]
210 100
205 90
185 50
180 40
175 30
170 20
165 10
160 0
5400 5450 5500 5550 5600 5650
Time [min]
48
Simulation Scenarios
As expected the missing load-share control possibility at the 2nd and 3rd section compressors result in a situation
where the A-train compressors (26-KA-002A, 26-KA-003A) takes most of the net-load, while a large part of the
gas is recycled in the B-train (Recycle valves are 35-45 % open for most of the simulation period).
In figure 5.37 and 5.38 the openings of the hot-gas by-pass valves and the openings of the anti-surge valves are
shown. Comparing the results of these graphs it is seen that, as the anti-surge valves reaches openings of 10 %,
which activates the hot-gas by-pass, process oscillations occur for a relatively short period of time. These
oscillations can be ascribed to a dynamic coupling between the anti-surge valve and the hot-gas by-pass valve.
After oscillating for approximately 10 minutes the system settles down with the operating points of the parallel
compressors moving away from each other.
The effect of the dynamic coupling problem on the actual volume flows to the compressors is difficult to see in
figures 5.37 and 5.38. An enlargement of the compressor inlet flows in the period of process oscillations can be
seen in figure 5.39 and 5.40.
630 178
625 176
26-KA-002A - Actual Volume Flow 26-KA-003A - Actual Volume Flow
26-KA-002B - Actual Volume Flow 26-KA-003B - Actual Volume Flow
620 174
Actual Volume Flow [m /h]
615 172
610 170
605 168
600 166
595 164
590 162
5440 5445 5450 5455 5460 5465 5470 5475 5480 5445 5450 5455 5460 5465 5470 5475
Time [min] Time [min]
Figure 5.39. 2nd section compressors inlet actual Figure 5.40. 3rd section compressors inlet actual
volume flows for the period of process volume flows for the period of process
oscillations. oscillations.
As seen from figure 5.39 and 5.40 the actual volume flows do at no point in time reach a flow where the
compressors are likely to surge. However, the dynamic coupling should be avoided, because under different
circumstances the surge flow might be reached, which could be devastating for the compressors.
5.2 Scenario B
The incoming well-fluid is 3-phase and hence the potential for slug flow is present. Especially severe slugging
pose a problem to the compression train because of severe oscillations in the gas flows can lead to pressure
oscillations in the gas compression trains. Due to the limited capacitance and limited possibility for capacity
control in the gas compression system, it is likely to be sensitive to severe slugging. Therefore, the performance
of the system upon slugging is investigated.
The main concern with slugs in the inlet well-fluid in regards to the gas compression system is that the
pressures in the compression system stay within the pre-defined pressure alarm settings given in table 5.12.
49
Simulation Scenarios
120000 120.00
Oil - Std Liq Vol Flow
Water - Std Liq Vol Flow
Gas - Std Gas Flow
100000 100.00
80000 80.00
60000 60.00
40000 40.00
20000 20.00
0 0.00
5465 5467 5469 5471 5473 5475 5477 5479
Time [min]
Three simulations have been run, with slug amplitudes of 30 %, 50 %, and 70 %, respectively.
The action initiated in the scenarios and the variables monitored are given in table 5.13.
Action Reduce inlet gas flow to 60 % of normal flow over a 30 minute period
Compressor inlet actual volume flows
Monitor
Compressor suction and discharge pressure
5.2.2 Results
Slug amplitude 30 %
Compressor inlet actual volume flow, and suction and discharge pressures plotted against time for simulations
with a slug flow amplitude of 30 % can be seen in figure 5.42 through 5.44.
50
Simulation Scenarios
26-KA-001A - Actual Volume Flow 26-KA-001B - Actual Volume Flow 26-KA-001A - Suction Pressure
26-KA-001B - Suction Pressure 26-KA-001A - Discharge Pressure 26-KA-001B - Discharge Pressure
4200 45
4000 40
3800 35
Actual Volume Flow [m /h]
3
3600 30
Pressure [barg]
3400 25
3200 20
3000 15
2800 10
2600 5
2400 0
5535 5540 5545 5550 5555 5560 5565
Time [min]
Figure 5.42. Actual volume flow, suction and discharge pressures for 1st section compressors. Slug
amplitude 30 % of normal flow and slug period 8 minutes.
26-KA-002A - Actual Volume Flow 26-KA-002B - Actual Volume Flow 26-KA-002A - Suction Pressure
26-KA-002B - Suction Pressure 26-KA-002A - Discharge Pressure 26-KA-002B - Discharge Pressure
800 120
750 100
Actual Volume Flow [m /h]
3
700 80
Pressure [barg]
650 60
600 40
550 20
500 0
5535 5540 5545 5550 5555 5560 5565
Time [min]
Figure 5.43. Actual volume flow, suction and discharge pressures for 2nd section compressors. Slug
amplitude 30 % of normal flow and slug period 8 minutes.
51
Simulation Scenarios
26-KA-003A - Actual Volume Flow 26-KA-003B - Actual Volume Flow 26-KA-003A - Suction Pressure
26-KA-003B - Suction Pressure 26-KA-003A - Discharge Pressure 26-KA-003B - Discharge Pressure
300 300
250 250
Actual Volume Flow [m /h]
3
200 200
Pressure [barg]
150 150
100 100
50 50
0 0
5535 5540 5545 5550 5555 5560 5565
Time [min]
Figure 5.44. Actual volume flow, suction and discharge pressures for 3rd section compressors. Slug
amplitude 30 % of normal flow and slug period 8 minutes.
Slug amplitude 50 %
Compressor inlet actual volume flow, and suction and discharge pressures plotted against time for simulations
with a slug flow amplitude of 50 % can be seen in figure 5.45 through 5.47.
52
Simulation Scenarios
26-KA-001A - Actual Volume Flow 26-KA-001B - Actual Volume Flow 26-KA-001A - Suction Pressure
26-KA-001B - Suction Pressure 26-KA-001A - Discharge Pressure 26-KA-001B - Discharge Pressure
4200 45
4000 40
3800 35
Actual Volume Flow [m /h]
3
3600 30
Pressure [barg]
3400 25
3200 20
3000 15
2800 10
2600 5
2400 0
5520 5525 5530 5535 5540 5545 5550
Time [min]
Figure 5.45. Actual volume flow, suction and discharge pressures for 1st section compressors. Slug
amplitude 50 % of normal flow and slug period 8 minutes.
26-KA-002A - Actual Volume Flow 26-KA-002B - Actual Volume Flow 26-KA-002A - Suction Pressure
26-KA-002B - Suction Pressure 26-KA-002A - Discharge Pressure 26-KA-002B - Discharge Pressure
800 120
750 100
Actual Volume Flow [m /h]
3
700 80
Pressure [barg]
650 60
600 40
550 20
500 0
5520 5525 5530 5535 5540 5545 5550
Time [min]
Figure 5.46. Actual volume flow, suction and discharge pressures for 2nd section compressors. Slug
amplitude 50 % of normal flow and slug period 8 minutes.
53
Simulation Scenarios
26-KA-003A - Actual Volume Flow 26-KA-003B - Actual Volume Flow 26-KA-003A - Suction Pressure
26-KA-003B - Suction Pressure 26-KA-003A - Discharge Pressure 26-KA-003B - Discharge Pressure
280 300
260 270
Actual Volume Flow [m /h]
240 240
3
Pressure [barg]
220 210
200 180
180 150
160 120
140 90
120 60
5520 5525 5530 5535 5540 5545 5550
Time [min]
Figure 5.47. Actual volume flow, suction and discharge pressures for 3rd section compressors. Slug
amplitude 50 % of normal flow and slug period 8 minutes.
Slug amplitude 70 %
Compressor inlet actual volume flow, and suction and discharge pressures plotted against time for simulations
with a slug flow amplitude of 70 % can be seen in figure 5.48 through 5.50.
54
Simulation Scenarios
26-KA-001A - Actual Volume Flow 26-KA-001B - Actual Volume Flow 26-KA-001A - Suction Pressure
26-KA-001B - Suction Pressure 26-KA-001A - Discharge Pressure 26-KA-001B - Discharge Pressure
4200 45
4000 40
3800 35
Actual Volume Flow [m /h]
3
3600 30
Pressure [barg]
3400 25
3200 20
3000 15
2800 10
2600 5
2400 0
5525 5530 5535 5540 5545 5550
Time [min]
Figure 5.48. Actual volume flow, suction and discharge pressures for 1st section compressors. Slug
amplitude 70 % of normal flow and slug period 8 minutes.
26-KA-002A - Actual Volume Flow 26-KA-002B - Actual Volume Flow 26-KA-002A - Suction Pressure
26-KA-002B - Suction Pressure 26-KA-002A - Discharge Pressure 26-KA-002B - Discharge Pressure
1000 140
120
900
Actual Volume Flow [m /h]
100
3
800
Pressure [barg]
80
700
60
600
40
500
20
400 0
5525 5530 5535 5540 5545 5550
Time [min]
Figure 5.49. Actual volume flow, suction and discharge pressures for 2nd section compressors. Slug
amplitude 70 % of normal flow and slug period 8 minutes.
55
Simulation Scenarios
26-KA-003A - Actual Volume Flow 26-KA-003B - Actual Volume Flow 26-KA-003A - Suction Pressure
26-KA-003B - Suction Pressure 26-KA-003A - Discharge Pressure 26-KA-003B - Discharge Pressure
300 300
270
250
240
Actual Volume Flow [m /h]
210
3
200
Pressure [barg]
180
150 150
120
100
90
60
50
30
0 0
5525 5530 5535 5540 5545 5550
Time [min]
Figure 5.50. Actual volume flow, suction and discharge pressures for 3rd section compressors. Slug
amplitude 70 % of normal flow and slug period 8 minutes.
Summary of results
For the slug scenarios simulated here it can be concluded that a situation with slugs with amplitudes of 30 %
and a period length of 8 minutes can be handled by the gas compression system. However, there are pressure
fluctuations with period lengths equal to the slugs on the discharge pressure of the 3rd section compressor of
245-259 barg, but it is within the PAL and PAH settings. On the 1st section compressor the pressure drops to
approximately 7 barg, which is below the PAL, when the gas flow is at its minimum. The flow also fluctuates
and on the 2nd and 3rd section compressors the surge control line is reached each slug period.
For the 50 % slug scenarios the pressure and flow fluctuations are increased. In addition, the problem of
dynamic coupling of the hot-gas by-pass valve with the anti-surge valves on the 2nd and 3rd section recurs.
However, at no point the surge flows are reached, but this is probably mainly due to the shortness of the period
length. If the slug periods were longer, the oscillations due to dynamic coupling would have time to amplify.
The pressures at the first and 2nd stage compressors stay within the boundaries of PAL and PAH, but on the 3rd
section compressors the discharge pressure reaches 270 barg, which is just above the PAHH. Hence, with the
current configuration this situation would be at the boundary for the capabilities for the gas compression
system.
In the 70 % slug scenario the effects seen in the 50 % scenario is greatly enhanced. The flow fluctuations due to
the dynamic coupled anti-surge and hot-gas by-pass valves are so significant that the surge flow is reached both
on 2nd and 3rd section compressors. Hence, the authenticities of the simulations are limited, but certainly it
illustrates the importance of dynamically de-coupling of the anti-surge and the hot-gas by-pass valves. A
maximum discharge pressure on the 3rd section compressors of 290 barg is reached which is well above the
PAHH of 269 barg.
A summary of the minimum and maximum pressures reached for suction and discharge are given in table 5.14.
56
Simulation Scenarios
Table 5.14. Minimum and maximum pressures reached for the 1st, 2nd, and 3rd section compressors.
Yellow values: PAL/PAH reached; Red values: PALL/PAHH values reached.
5.3 Scenario C
The overall idea with the OGPU has been to design it to be generic in the meaning of it to be all-round.
Therefore it can be anticipated that the gas to be handled will vary in composition and moleweight. From the
discussion of compressor performance in section 2.4 this significantly alter the performance of the compressors
as less head is required to compress a denser gas and more head is required to compress a leaner gas.
In order to investigate the effects on the total compression system two scenarios have been setup where the
composition is changed to get leaner and denser gas, respectively. The compressor performance curves used for
these simulations are unaltered. Thus, there are no corrections made for the stage mismatching effects that are
likely to be present when changing the composition. An optimal situation would be to acquire compressor
curves made for the composition used in these scenarios. However, it is believed that the results obtained from
the simulations give a reasonable picture of the overall performance of the compression system upon a gas
composition change.
Only the inlet gas composition is changed and hence the inlet flow-rate is similar to that used in the base case.
Leaner Gas
The inlet compositions used for the leaner gas case are given in table 5.16.
57
Simulation Scenarios
Denser Gas
The inlet compositions used for the denser gas case are given in table 5.17.
58
Simulation Scenarios
5.3.2 Results
The results given in the following are found after the model has been allowed to reach steady state and are
reported as average values found for a 20 minute period (simulation time). It could be argued that a steady state
model would do the job, but that is not the case. This is because in the HYSYS DynamicsTM model the
pressure/flow equations are solved based on resistance equations, which is not the case in steady state modeling
where pressure/flow relations are specified by the user. A more thorough discussion can be found in appendix
A.
The composition changes applied to the model result in the following moleweights at the compressors as given
in table 5.18.
Compressor section
Case
1st 2nd 3rd
Lean gas 21.39 20.70 20.70
Normal gas 23.63 22.71 22.71
Dense gas 25.91 24.70 24.70
Hence the deviations in moleweight from the base case are approximately ±2 g/mole for all compressor sections.
In table 5.19 the resulting average actual volume flows at the compressor suction are given.
Compressor section
Case
1st 2nd 3rd
Lean gas 2667 m3/h *584 m3/h 190 m3/h
Normal gas 3347 m3/h 667 m3/h 199 m3/h
3
Dense gas 3803 m /h 770 m3/h 209 m3/h
As seen from table 5.19 the anti-surge control line for the lean gas case at the 2nd section compressors have
moved down to a flow of 584 m3/h, which is just 6 % above the presumed surge flow. In itself this is not
troubling but it shows that the coordinate system used for surge control is not truly invariant to changes in the
process conditions.
In general the flows are lower for the leaner the gas, which is due to the fact that more head is needed to obtain
the same compression ratio for lean gas as compared to heavy gas.
Compressor suction pressures for the three setups are given in table 5.20.
Compressor section
Case
1st 2nd 3rd
Lean gas 11.4 barg 43.4 barg 112.5 barg
Normal gas 8.9 barg 36.3 barg 101.1 barg
Dense gas 7.9 barg 31.4 barg 89.9 barg
The suction pressures for the lean gas situation need to be greater because the compressors deliver lower
pressure ratios as the moleweight is decreased. The pressures in the lean gas case are all close to the PAH
59
Simulation Scenarios
settings given in table 5.12, whereas the pressures in the dense gas case all are close to the PAL settingns. As
such these cases may define the boundaries of obtaining normal operation.
Finally the discharge pressures of the compressors are given in table 5.21.
Table 5.21.
Compressor section
Case
1st 2nd 3rd
Lean gas 44.2 barg 112.9 barg 251.4 barg
Normal gas 37.4 barg 101.5 barg 251.5 barg
Dense gas 32.7 barg 90.3 barg 251.5 barg
Summary of results
In summary it has been shown that a gas moleweight deviation of ±2 g/mole at the compressors can be handled
by the gas compression system under normal operating conditions. However, it is indicated that due to the
inter-section pressures reaching the PAL/PAH levels a moleweight deviation of ±2 g/mole seems to be the limit
of normal operation.
60
Discussion
6 Discussion
In this section issues related to the current project are discussed. The discussion is divided into two sections
Modifications to the present system are proposals of initiatives that can enhance the operability of the gas
compression system on the OGPU.
Future work proposals are given both with regard to the OGPU gas compression system, but also with regard to
working with models in HYSYS DynamicsTM.
61
Discussion
Short explanations of the above listed proposals are given in the following.
62
Conclusion
7 Conclusion
The primary objective of this project has been to create a realistic model of the oil and gas processing
equipment on an Oil and Gas Production Unit (OGPU). The model has been created using the general purpose
simulation tool HYSYS DynamicsTM. Special attention has been given to the modeling of the gas compression
train and to the setup of the compressor control system.
The model has been used to investigate the performance of the gas compression system at off-design conditions.
Through the simulations it has been shown that the system can handle sluggish flow at amplitudes of 30 % of
normal flow at a period length of 8 minutes. At larger amplitudes the system has proven unstable, especially
because dynamic de-coupling of anti-surge valves and hot-gas by-pass valves. The problem is likely to be solved
by applying loose control to the hot-gas by-pass controllers.
With the current configuration it has been found that it is not possible to share the compression load between
the parallel coupled compressors at the 2nd and 3rd section of the compression train. Therefore, as the anti-surge
control is activated a situation occurs where one compressor train takes a larger quantity of the net-load while
the other recycles. In order to avoid the situation and thereby achieve a less power consuming compression two
suction throttling valves could be implemented at the 2nd section compressors.
The OGPU has been designed based on a specific well-fluid composition representative of the North Sea region.
However, as the has not been designed for a specific field, but rather as a generic OGPU it is likely that it is
going to operate at a field with a well-fluid composition differing from the one used as design basis. Therefore,
the model has been used to investigate how the system will respond to changes in the inlet gas composition. It
has been found that gas compositions with deviations in the moleweight at the compressors of ±2 g/mol are
close to the boundary of what the system can handle with the current configuration.
A secondary objective is to contribute to the evaluation of HYSYS as it is a candidate to substitute currently
used simulation tools at Aalborg University Esbjerg. The project set an example of the possibilities of using
HYSYS DynamicsTM as a process simulation tool. The extensiveness of the possibilities available in the program
– especially the possibility of implementing self-made pieces of code – makes it well-suited for use on the
master level educations. In addition, it can be used for steady state as well as dynamic simulations, which is not
the case with the presently used process simulation tool.
63
References
References
API Standard 617. Axial and Centrifugal Compressors and Expander-compressors for Petroleum, Chemical, and
Gas Industry Services; Downstream Segment 2002, 7th ed.
Aspen COMThermo, Reference Guide. Aspen Technology, Inc., Burlington, Massachusetts, 2007.
Aspen HYSYS, Dynamic Modeling. Aspen Technology, Inc., Burlington, Massachusetts, 2007.
Aspen HYSYS, Operations Guide. Aspen Technology, Inc., Burlington, Massachusetts, 2007.
Aspen HYSYS, Simulation Basis. Aspen Technology, Inc., Burlington, Massachusetts, 2007.
Aspen HYSYS, Tutorials and Applications. Aspen Technology, Inc., Burlington, Massachusetts, 2007.
Aspen HYSYS, User’s Guide. Aspen Technology, Inc., Burlington, Massachusetts, 2007.
AspenTech HYSYS, Version 2006.5 (build 21.0.2.6924); Aspen Technology, Inc., Burlington, Massachusetts,
2007.
Bloch, H. P. A Practical Guide to Compressor Technology; 2nd ed.; John Wiley & Sons, Inc., Hoboken, New
Jersey, 2006.
Gresh, M. Theodore. Compressor Performance: aerodynamics for the user; 2nd ed.; Butterworth-Heinemann:
Woburn, Massachusetts, 2001.
Hanlon, P. C. Compressor Handbook; 1st ed.; McGraw-Hill: New York, 2001.
ISO 5167-1. Measurement of fluid flow by means of pressure differential devices, Part 1: Orifice plates, nozzles
and Venturi tubes inserted in circular cross-section conduits running full; Measurement of fluid flow in closed
conduits 1991, 1st ed.
Lüdtke, K. Aerodynamic Stage Mismatching – The Key to Understand Multistage Process Centrifugal
Compressor Behavior; VDI Berichte, no. 1425, Düsseldorf, Germany, 1998.
Lüdtke, K. H. Process Centrifugal Compressors: Basics, Function, Operation, Design, Application; 1st ed.;
Springer-Verlag Berlin Heidelberg New York, 2004.
Microsoft Office Excel 2003, Version 11.5612.5606; Microsoft Corporation.
Microsoft Visual Basic 6.3; Microsoft Corporation.
Rammler, R. Considerations for advanced centrifugal compressor control; ISA Transactions, Vol. 33, p. 153-157,
Houston, USA, 1994.
Sandler, S. I. Chemical, Biochemical, and Engineering Thermodynamics; 4th ed.; John Wiler & Sons, Inc.:
Hoboken, New Jersey, 2006.
Seborg, D. E., Edgar, T. F., Mellichamp, D. A. Process Dynamics and Control; 2nd ed.; John Wiley & Sons, Inc.,
Hoboken, New Jersey, 2004.
Shinskey, F. G. Process Control Systems: Application, Design, and Tuning; 4th ed.; McGraw-Hill, New York,
1996.
Svrcek, W. Y., Mahoney, D. P., Young, B. R. A Real Time Approach to Process Control; 2nd ed.; John Wiley &
Sons Ltd, Chichester, West Sussex, England, 2006.
Tham, M. Discretised PID Controllers – Part of a set of study notes on Digital Control. Chemical Engineering
and Advanced Materials; University of Newcastle upon Tyne; Web-site visited June 4, 2008.
http://lorien.ncl.ac.uk/ming/digicont/digimath/dpid1.htm#Positional
Turton, R. K. Principles of Turbomachinery; 2nd ed.; Chapman & Hall, London, UK, 1995.
64
Appendix A – Simulation in HYSYS
A Simulation in HYSYS
This appendix is an introduction to simulation in HYSYS, which is a general purpose process simulation tool
developed for use in the oil and gas business. However, it can be used for a wide variety of applications not
related to oil and gas as well.
The present appendix is meant as an introduction to the possibilities in using HYSYS. It is not a step-by-step
tutorial but rather a short compilation of some of the possibilities of simulating in HYSYS. For more detailed
information and access to tutorials the program documentation should be consulted (attached on CD-ROM).
The possibilities for simulations in HYSYS described in this appendix are available if using the licenses in the
AspenTech University Package for Process Modeling. However, the particular programs and plugins must be
installed to be available.
Two general modes of simulating in HYSYS are steady state and dynamic simulation. The basis for the two
types of simulations is equal but in dynamic mode time and volume are used in addition to the mass and energy
balances solved; i.e. the accumulation term is included.
The subjects discussed in this appendix are
• Program Structure
• Steady state simulation
• Dynamic simulation
65
Appendix A – Simulation in HYSYS
The first two tabs – Components and Fluid Pkgs – are essential to any simulation be it steady state or dynamic.
Component selection
Components can be added to the Master Component List by pressing the View… button as seen on figure A.1.
By pressing Add multiple component lists can be made which can be an advantage in simulations where
separate systems have different combinations of components. In case of multiple component lists the Master
Component List contains all the components used in the component lists. Standard components can be selected
from a large database.
In addition to the standard components available in the HYSYS database it is possible to generate pseudo
components in the Hypotheticals tab. These are often used for separation processes in the oil and gas industry.
The data necessary for creating a pseudo component is Normal Boiling Point (NBP), Molecular Weight (MW),
Std. Liquid Density, Critical Temperature (Tc), Critical Pressure (Pc), Critical Volume (Vc), Acentricity. It is,
however, possible to let HYSYS estimate one or more of these properties. The minimum amount of data
necessary is NBP or Std. Liquid Density and MW.
66
Appendix A – Simulation in HYSYS
A.2.1 Specifications
Applying specifications to the model is frequently done by specifying the conditions at the inlet and outlet of
the equipment. However, it is also possible to define a pressure drop over a valve or a certain power input to
and efficiency of a pump, etc. The only rule is that the number of specifications must match the number of
equations for a certain operation; i.e. it must neither be underspecified nor overspecified.
67
Appendix A – Simulation in HYSYS
The starting point of stream specification is to define a composition, flowrate, pressure, and temperature of one
stream in the PFD. Because HYSYS is event driven, it calculates all other properties of the stream when these
four specifications have been applied. After specifying one material stream specifications must be given to the
unit operations and/or material and energy stream throughout the PFD. Whenever a material stream – or a unit
operation – is fully specified HYSYS performs the calculations and it turns dark blue.
The most common error encountered when simulating in HYSYS steady state is probably that equipment is
over specified. User specified values are always written in blue and thus if an over specification error occur
trouble-shooting consist in examining the specifications applied to the model.
A.2.2 Equipment
The configuration of common pieces of equipment is described in the following.
Valves
In steady state, valves are simply means of reducing pressure. As mentioned before the pressures of the material
streams connected to the valves can be specified (thus defining the valve pressure drop) or a specific pressure
drop can be defined in the valve Design/Parameters tab. In the Rating/Sizing tab the valve can be sized giving
a Cv value. It is important to note that this value is not used for any calculations when in the steady state mode
of HYSYS. It is merely a possibility to calculate an appropriate size of the valve.
Heat exchangers
The simplest forms of heat exchangers are heaters and coolers where an effect can be defined to give a
temperature change or vice versa.
A more realistic approach is to simulate heat exchangers as e.g. shell and tube or plate heat exchangers. Several
possibilities for configuring the heat exchangers exist. The simplest is to configure it as a constant UA
exchanger by specifying UA. However, if data is available it is possible to apply detailed information of e.g.
length, diameter, and pitch of tubes, details of materials used etc. From these details HYSYS calculates the Area
and Heat Transfer Coefficient.
The default method for heat exchanger design in HYSYS is the end-point model where the total heat
transferred, Q, is calculated by
Q = U ⋅ A ⋅ ΔT LM ⋅ Ft (A.1)
Where U: Overall heat transfer coefficient
A: Surface area available for heat transfer
ΔT LM : Log-mean temperature difference (LMTD)
Ft : LMTD correction factor
68
Appendix A – Simulation in HYSYS
The end-point model should be used when there is no phase change and when Cp’s of both fluids are
approximately constant. If phase changes occur on either side of the exchanger the weighted model should be
used. Here the heating curves are broken into intervals and a heat balance performed along each interval.
Yet another possibility is to use one of the advanced heat exchanger simulation programs supplied by
AspenTech as plugins in HYSYS. Each plugin must be activated on the Design/Parameters tab of the specific
equipment. The plugins available for the different types of heat exchangers are given in table A.1. It should be
noted that these plugins only can be used for steady state simulation in HYSYS.
Pipes
Pipes can be applied to a model as a Pipe segment which can be used for both gas and liquid flow or as a Gas
pipe which uses models only applicable to pure gas flow.
The default correlation for pressure drop calculation in the Pipe segment is the Beggs and Brill correlation
because it can model liquid holdup, vertical flow, horizontal flow, and predict the flow regime. The other
methods available all have a lack of at least one of these four modeling abilities. The OLGAS models are not
available using the AspenTech University Package.
The Gas pipe is originally meant for dynamic simulations but is available for use in steady state as well. The
only advantage of using such pipes in steady state is that the maximum Mach No. is reported.
Separators
Separators can either be 2-phase or 3-phase. Tanks are simulated exactly the same way as 2-phase separators.
The basis of the separators is the calculation of the thermodynamic equilibrium. Separators can have multiple
inlets and it is possible to apply energy streams to the separators. For 3-phase separators the bottom outlet is the
heaviest liquid, the middle outlet is the light liquid, and the top outlet is the gas.
It is possible to simulate imperfect separations by application of the Carry over model available.
Columns
Columns available in HYSYS are
The setup of these except for the component splitter and short-cut distillation column are made by following
the steps provided when opening the unconfigured column.
It can be a difficult task to select an appropriate number of stages and reflux ratio for a distillation column. An
approach to this is to use the short-cut column where a desired split can be defined and hereafter a minimum
69
Appendix A – Simulation in HYSYS
number of stages will be calculated for a given reflux ratio. The calculated values can hereafter be applied to an
ordinary distillation column that performs a more rigorous calculation.
The component splitter can be used to remove certain components from a stream. Thus, it is merely a kind of
pseudo splitter as it does not reflect a real process.
Logical operators
Commonly used logical operators in HYSYS steady state are Adjust, Set, and Recycle.
The Adjust operation can be used to obtain a required value or specification by iteration. Thus, an independent
value is changed until the dependent variable reaches the required value.
The Set operation can be used to transfer a process or stream specification from one unit or stream to another.
The Recycle operation must be used whenever there is a recycle in the simulation for example when upstream
material is mixed with downstream material or vice versa. The Recycle operator creates a break in the
simulation allowing calculation of the entire flowsheet to be based on previously guessed values. Afterwards the
guessed values are updated and the calculation of the flowsheet is repeated. This iteration process is continued
until inlet and outlet of the Recycle operator are equal.
Spreadsheet
It is possible to insert a spreadsheet into the simulations. These can be used to gather up relevant process data in
lists. From the spreadsheet it is also possible to import the user specified variables and changing the from the
spreadsheet.
70
Appendix A – Simulation in HYSYS
applications it can be a good idea to reduce the timestep, especially if the accuracy of the absolute values
coming out of the model has great importance.
For most applications the default execution rates should not be altered.
A.3.3 Equipment
As mentioned previously the accumulation term is included in the mass and composition balances in dynamic
simulation. Hence, volumes must be entered for all equipment, unless it is regarded unimportant (this could be
the case for the hold-up volume of e.g. a heat-exchanger or a valve).
In the following the application of common pieces of equipment to a HYSYS DynamicsTM model will be
described with emphasis on the differences to steady state simulation.
⎛ 59.64 p2 ⎞
Qvap = υfracfac ⋅ 1.06 ⋅C g ⋅ ρ ⋅ p1 ⋅ sin ⎜⎜ 1− ⋅ cp fac ⎟⎟ (A.2)
⎝ C1 p1 ⎠
Cg Km 0.4839 Cp
Where C1 = C1 = cp fac = ⎛ γ ⎞
γ=
Cv 0.001434 ⎜ ⎟ Cv
⎛ 2 ⎞ ⎝ γ −1 ⎠
1−⎜ ⎟
⎝1+γ ⎠
Cv : Fluid flow sizing coefficient in USGPM (US gallons per minute)
Cg : Gas flow sizing coefficient in USPGM
71
Appendix A – Simulation in HYSYS
In addition to the valve characteristics it is possible to specify the actuator type as instantaneous, linear, or first
order reacting. It is also possible to specify a maximum opening rate etc. Another option is to specify a
minimum and/or maximum position of the valve/actuator.
72
Appendix A – Simulation in HYSYS
Heat exchangers
Heat exchangers are simulated as constant UA exchangers and the heat transfer is calculated by equation A.1.
Since UA is a function of flow it is possible to enter a reference flow and the UA used is then calculated by
0.8
⎛ m& ⎞
U ⋅ Aused = U ⋅ ASpecified ⋅ ⎜ current ⎟ (A.4)
&
⎝ mreference ⎠
Pressure drop across the heat exchanger is best modeled by specifying a k-value from which the pressure drop is
calculated as
m& 2
Δp = (A.5)
ρ ⋅k 2
Where m& : Mass flow-rate
k: Constant for frictional pressure loss
It is not possible to use the heat exchanger plug-ins in the dynamic simulation mode.
Pipes
Pipes are simulated as in steady state mode. Even though it is possible to simulate slugs in steady state it is not a
possibility in dynamic mode.
Long pieces of pipeline are best simulated by using either Gas Pipe or Pipe Segment because in these models
the pipeline can be broken down into a number of segments. AspenTech recommends that each pipe-segment
has a pressure drop of less than 10 % of the absolute pressure.
For smaller pipelines such as those between processing equipment it can be advantageous to apply a valve and
select the Disable Valve (Pipe Only) in the Dynamics/Pipe tab. By this method the pipeline is modeled by a
simple pressure flow relation. However, the hold-up volume is not automatically calculated but has to be
entered manually. The advantage of this is that it is possible to enter the length as an equivalent length and the
hold-up volume as the actual hold-up volume. Another advantage is shorter computation time which is a huge
advantage for large complex systems.
Separators
The modeling of separators does not differ significantly from steady state other than the necessity of defining a
volume. It should be noted that the placement of nozzles on the separator dictates which phase is drawn from
the separator.
Logical operators
In dynamic mode the logical operator Adjust is typically replaced by a PID Controller, however, other
controller types are available.
The Recycle operator can be used but it is not necessary. The set operator can be used in the same manner as in
steady state mode, however, is may not always resemble an actual process. In most cases application of a PID
Controller better resemble actual processes.
A PID controller in HYSYS solve the characteristic control equation
Kc de (t )
OP (t ) = OPss + K c ⋅ e (t ) +
Ti ∫ e (t ) dt + K c ⋅Td
dt
(A.6)
73
Appendix A – Simulation in HYSYS
Kc : Controller Gain
Ti : Integral time
Td : Differential time
In HYSYS the PID control equation can be solved by use of two basically different algorithms. These are
In the velocity (or differential) form the discretized controller equation is given by
⎪⎧ ⎡e (t ) − 2e (t − 1) + e (t − 2 )⎤⎦ ⎪⎫
e (t ) ⋅ h + Td ⎣
1
u (t ) = u (t − 1) + K c ⎨e (t ) − e (t − 1) + ⎬ (A.7)
⎪⎩ Ti h ⎪⎭
⎪⎧ ⎡e (t ) − e (t − 1)⎦⎤ ⎪⎫
e ( i ) + Td ⎣
n
h
u (t ) = K c ⎨e (t ) + ∑ ⎬
⎩⎪ Ti i=1 h ⎭⎪
Thus, by using the positional form of discretization the integral term is summed up continuously, as opposed to
the velocity form where the integral term is calculated for each sampling period. A side effect of the
continuous summation of errors in the positional form algorithm is called integral windup. The effect is best
described by looking at a typical response for a PID controller upon a SP change and comparing it with the
response when integral windup occurs.
Signal Signal
PV PV
B
A
SP Typical Response SP Typical response
without with
Integral Windup Integral Windup
Time Time
As seen in figure A.3 the integral wind-up effect tends to prolong the period of overshoot that occurs upon a SP
change. This is because the positional form algorithm sums up all previous errors and therefore it is the first
overshoot period basically continues until the areas A and B are equal. This is not the absolute truth because
other corrective action is also taken but it illustrates the effect of integral windup.
74
Appendix A – Simulation in HYSYS
Therefore, the positional form algorithm should, as a general rule, not be used whenever the controller is setup
with integral action. The integral wind-up phenomenon is also the reason that the velocity form algorithm is
set as default in HYSYS.
75
Appendix B – HYSYS DynamicsTM PFD for the OGPU
Figure B.1. Process Flow Diagram of the OGPU as modeled in HYSYS DynamicsTM.
76
Appendix C–Anti-surge controller tuning
77