ModellingofSevereSlugandSlugControlwithOLGASPE84685 PDF

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 31

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/240780131

Modeling of Severe Slug and Slug Control With OLGA

Article  in  SPE production & operations · August 2006


DOI: 10.2118/84685-PA

CITATIONS READS
8 1,466

3 authors, including:

Kelvin Ling John-Morten Godhavn


Curtin University Norwegian University of Science and Technology
5 PUBLICATIONS   193 CITATIONS    60 PUBLICATIONS   1,256 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Intelligent Drilling -- Automated Underbalanced Drilling Operations View project

Controlled Mud Level View project

All content following this page was uploaded by John-Morten Godhavn on 08 August 2014.

The user has requested enhancement of the downloaded file.


Modelling of Severe Slug and Slug Control with OLGA
SPE Number 84685
Mehrdad P. Fard John-Morten Godhavn
Hydro, Oil & Energy Statoil ASA
Brage Petroleum Technology R&D, Process Control
Sandslivn 90 PO Box 7190 Arkitekt Ebells vei 10, Rotvoll
N-5020 Bergen, Norway N-7005 Trondheim, Norway
Phone: +47 99306939 Phone: +47 90974301
Fax: +47 55996928
E-mail: Mehrdad.Fard@Hydro.com E-mail: jmgo@statoil.co
Svein Ivar Sagatun, SPE
Hydro, Oil & Energy
Troll Petroleum Technology
PO Box 7190 - N 5020 Bergen, Norway
Phone: +47 55 99 5918
Fax: +47 55 99 6639
E-mail: Svein.Ivar.Sagatun@hydro.com

Summary

With the recent developments in the oil and gas industry where the number of satellite fields
are increasing, the industry faces major challenges. A mixture of gas and oil is transported in
multiphase pipelines along the bottom of the sea from the wells and up through the riser to
the oilrig. The development of unstable flow in multiphase pipelines is a major and expensive
problem. The irregular flow results in poor oil/water separation, limits the production capac-
ity and causes flaring. Handling slug flow, thus, has become very important with regards to
increase the production rate and avoiding possible equipment damages.
In this paper we have developed a model of a physical process that generates slugs. The
model has been verified against experimental data. Several control strategies have been tested
on the model and the simulation results are presented. It is shown by simulation of multi-
phase flow that the unstable slug flow can be stabilized by feedback control. In addition, in-
teresting and important phenomena such as inverse response of the pressure at the top and
asymmetric response of the pressure at the bottom have been revealed and explained.

1
1. Introduction may cause poor separation and occasionally fluid

Multiphase pipelines connecting remote well- flooding. Oscillations in the separator pressure

head platforms and subsea wells are a common can result in increased flaring which is environ-

feature of offshore oil production. Recent devel- mentally benign and costly due to CO2 taxes.

opments indicate that this trend will be followed The degree of slugging depends on various fac-

in the future. In addition, the feasibility of using tors. The most important are pipeline pressure,

long-distance tie-back pipelines to connect sub- pipeline topology and production rate. Terrain

sea processing units directly to on-shore process- slugging is most likely to occur at low rates with

ing plants makes it likely that these will also be a low pipeline pressure towards the end of a

deployed in the future. These trends in offshore well’s lifetime.

oil production face a major challenge. There have been reported considerable opera-

The slug phenomenon in multiphase flow lines tional problems and equipment damages due to

with both gas and liquid hydrocarbons has slugs. Some of the typical problems encountered

gained increased interest in the latest years. at various platforms are:

Slugging can be characterized as either hydrody- Large disturbances in the separator train, causing

namic or terrain slugging. Hydrodynamic slugs Poor separation, high water-in-oil percentage in

are built in horizontal parts of the pipeline when the oil outlet.

liquid and gas velocities are different. These Varying water quality at the separator water out-

slugs are usually short with higher frequency. lets, which in turn causes difficulties for down-

The inlet separator will in most cases handle stream water treatment system.

these slugs with less difficulties, since the Large and rapidly varying compressor operation,

amount of liquid is small compared to the vol- due to pressure variations.

ume of the separator. On the contrary a terrain Large pressure fluctuation, which causes reduc-

slug can contain a lot of liquid and represents a tion in lifting capacity of the wells.

great challenge to the downstream processing Various attempts have been made to avoid the

system. Level and pressure variations in the 1. problems associated with slug flow, [1-5]. Some

stage separator can be propagated in the down- of these are design related such as installing slug

stream process. Variations in the separator level catchers on-shore or increasing the size of the

2
first stage separators to provide the necessary
The process to be modelled is an experimental
buffer capacity. However, such solutions are not
set-up at the SINTEF Petroleum Research Multi-
appropriate due to high costs and needs for larger
phase Flow Laboratory [1] at Tiller outside
space capacity.
Trondheim. The gas SF6 and the liquid Exxsol
This paper describes the Olga1 simulation model
D80 are mixed in a 231m, 3” closed loop with a
of a physical process and compares the simula-
vertical riser for circulation of oil and gas. The
tion results with experimental data. Furthermore,
process equipments between the inlet and outlet
this paper shows how the Olga model can be
of the loop are shown in Figure 1. The first 100
used in the control design phase prior to testing.
meters have a -0.1° declination, then a 180° hori-
Phenomena such as inverse response of the pres-
zontal U-turn. The diameter of the U-turn is
sure at the top and asymmetric response of the
3.5m, corresponding to a length of 70 diameters.
pressure at the bottom have been revealed and
After the U-turn, the pipe is declined -0.7° for
explained. These phenomena play crucial rules in
about 105m, and finally a 15m vertical riser. The
the design of controller strategies for stabilizing
riser ends in a double bend, where the flow is
the unstable slug flow. This paper is organized as
directed downwards into an 8” vertical drop-leg.
follows. The test loop, being modelled, is pre-
The vertical drop leg ends in a gas-liquid separa-
sented in Section 2. Section 3 is devoted to the
tor, where the gas is drawn into a de-mister to
modelling and verification issues. Some interest-
remove droplets and then into the compressor.
ing and important observations are discussed in
The oil is drained to the horizontal separator, and
Section 4. In section 5, slug suppression by aims
recycled through the oil pump. At the inlet sec-
of active control has been discussed and several
tion the gas is mixed with the oil through a 45°
control strategies are tested on the developed
downward inclined pipe. The separated phases
model. Finally, some concluding remarks are
are fed to the single-phase velocity measurement
made in Section 6.
stations and routed through flow meters by

manually operated valves. The oil and gas then

passes through a 7m long flexible (rubber) pipe


2. Loop Description
section, and on to the initial -0.1° section. The
1
Olga 2000 is a state-of-the-art multiphase flow
simulator available from Scandpower. scientific instrumentations are located along the

3
entire loop, but mostly along the last 100-meter • 105(m) pipe inclined –0.7 deg.

section and in the riser. The hold-up is measured • 15(m) vertical riser

by means of seven single-energy narrow-beam The x-y coordinates of the pipelines are shown in

gamma densitometers distributed along the pipe, Table 1. The topology of the pipeline is shown in

with the two last ones in the riser. Figure 2.

The frequency density responses presented in The pipeline has been modelled as 3” pipe with

this paper are generated using the FFT function inner diameter of 69mm. An internal roughness

of Matlab. of 3e-5 was assumed.

3. Olga Modelling of Tiller Loop The pipeline is divided into pipe elements of

The main objective of this section is to develop constant cross sectional area with various

an Olga model that captures the main physical lengths. Each pipe is subdivided into sections.

mechanism of the slugs generated in the Tiller The number and the length of sections vary for

loop. The model developed in this section shows each pipe element. The pipeline model consists

some discrepancies such as offset in valve open- of 7 pipe elements with 50 sections.

ing and over calculation of the main frequency

component in the top pressure. The model is in- 3.2. Boundary Conditions. The pressure at

tended to be used in the control design stage of the inlet separator is assumed to be constant at

the study, therefore it is important that the flow 1.2 [bar]. Therefore, the outlet boundary condi-

and pressure oscillation frequencies and ampli- tion is modelled as a constant. In order to avoid

tudes are close to the real data. This point has numerical disturbances introduced in the model

been confirmed in our verification study. The an extra horizontal pipe is added to the top of the

terrain slug mechanism of slug tracking module riser.

of Olga 2000 version 2.03 is used in the simula- The pipeline inlet is modelled as a source with a

tions. The default value of delay constant is used. specified total mass flow rate. This module en-

3.1. Loop Topology. The loop has been sures a constant mass flow rate into the loop. The

modelled with the following topology: simulations were performed by a constant mass

• 100(m) pipe inclined -0.1 deg. flow rate of 0.8 [Kg/s]. The gas mass fraction is

• 10(m) flexible pipe for U-turn set to 0.0023.

4
3.2. Model Verification. A simulation is Figure 7. It can be seen that there is a discrep-

run to verify the model against the experimental ancy in the valve opening. The mean value of the

data. The simulation is run with fully open valve valve opening from measurements is about 24%,

as in the experiment. The comparison of the with automatic control. However, the mean value

pressures at the top and the bottom of the riser is of the simulated valve opening is about 46%.

shown in Figure 3 and Figure 4. It is shown that 4. Observed Phenomena

the frequency and the amplitude match very well The following observations in the simulations are

in the case of pressure at the bottom of the riser. in accordance with what have been reported from

However, there is some discrepancy between the corresponding experiments in [6]. These behav-

simulated and measured pressure at the top of the iour may have a more significant impact on the

riser. From frequency density, upper plot, in responses in a larger scale system, where the

Figure 3 we can see that the main frequency length of the riser is significantly larger.

component of top pressure response computed by 4.1. Asymmetric Pressure Response.


Olga is almost twice the measured top pressure. An interesting observation was made with re-
For further verification of the developed model gards to the step response of the pressure. A
for dynamic responses, a one and half hour sce- small increase in the valve opening during stable
nario is simulated. The same scenario is run as an flow, see Figure 8, results in a much faster pres-
experiment in order to collect data for verifica- sure reduction in the bottom of the riser than the
tion of the model, see Table 2. pressure increase one gets with a similar small
The controller parameters for both simulation decrease in the valve opening, see Figure 9. A
and experiment are given in Table 3. decreased valve opening increases the amount of
Note that the valve opening in the Olga 2000 liquid in the riser. The pressure build-up is lim-
varies in the [0,1] intervals while in the experi- ited by the liquid rate into the riser, and therefore
ment the valve opening can vary in the [0,100] slow. An increased valve opening reduces the
intervals. This is the reason why the gains are amount of liquid in the riser. The liquid is re-
different. placed by (compressed) gas in the pipeline that
The responses are compared and shown in Figure will expand in the riser, and therefore fast. The
5 and Figure 6. The valve openings are shown in asymmetry cannot be handled properly with lin-

5
ear controllers. The controller must be tuned with during stable flow (almost constant pressure both

respect to the fastest response. This means e.g. at the top and at the bottom of the riser) gives a

that a pressure controller must be evaluated when pressure fall in the bottom of the riser as ex-

the pressure reference is decreased. The per- pected. However, we observed a temporary pres-

formance with an increased pressure reference sure increase in the top of the riser, before the

will then be slow. top pressure was decreased to a new stationary

value, see Figure 8. Similarly, a small closure of


4.2. Inverse Response at Top Pressure.
the valve gives the expected pressure increase in
An alternative to using bottom pressure meas-
the bottom and a more surprisingly pressure de-
urements is to use pressure measurements up-
crease at the top before the expected stationary
stream the topside choke (top pressure). This
increase, see Figure 9.
alternative has the advantage of being independ-
A decreased valve opening gives a reduced mass
ent of underwater instrumentation. Bottom pres-
rate through the valve and out of the riser. And
sure measurements are in most cases less avail-
with a constant rate into the riser, liquid will be
able and more expensive than topside measure-
accumulated in the riser and give an increased
ments. However, the top pressure behaves differ-
bottom pressure. The relative velocity between
ent from the bottom pressure. The top pressure is
the gas and liquid phases (slip velocity) in the
given by the choke downstream pressure (separa-
top of the riser increases, resulting in less liquid
tor pressure) and the differential pressure over
(mass) through the valve. A reduced mass rate
the choke. The latter is again given by the rate
through the valve results in a reduced differential
flowing through the valve while the bottom pres-
pressure across the valve. The same reduction in
sure is given by the top pressure plus hydrostatic
the differential pressure across the valve will also
pressure given by the amount of liquid in the
take place at the top pressure, since the separator
riser.
pressure is constant. Liquid is now accumulated
We observed that the top pressure has an inverse
in the riser, since less liquid flows out and the
response compared to the bottom pressure on
flow into the riser is constant. The liquid build-
steps in the valve opening (open loop experiment
up and bottom pressure increase is the first part
– no feedback). A small opening of the choke
of a pressure oscillation towards a new equilib-

6
rium. This can result in a riser slug if the valve 5.1. Single Loop Bottom Pressure
opening change is sufficiently large. Similarly,
Control. This control scheme uses feedback
an increased valve opening gives an increased
from pressure at the bottom of the riser. The bot-
mass rate out of the riser, reduced relative veloc-
tom riser pressure will increase when the amount
ity between gas and liquid, increased differential
of liquid in the riser increases. The idea is to sta-
pressure across the valve, increased top pressure,
bilize the amount of liquid in the riser by using
less liquid in the riser and reduced bottom pres-
the topside choke, and in this way avoid slugs to
sure.
build up. The control structure is shown in
This behaviour of the top pressure imposes limi-
Figure 10.
tations on the performance of the system from a
The controller parameters are shown in Table 4.
control point of view. This issue will be ad-
The simulation starts with fully open valve. The
dressed later in the paper.
slug period is about 204 seconds. The controller
5. Slug Control using Olga/Matlab
is then switched on after 350 seconds. The set-

Toolbox point is reached after almost 10 minutes. The top

This section is dedicated to the control issues of and bottom pressures are shown in Figure 11,

the model developed in Section 3. Scandpower, while the volumetric flow rate and the valve

in corporation with Norsk Hydro ASA, has de- opening are shown in Figure 12.

veloped an Olga/Matlab toolbox, [7]. This tool- 5.2. Cascade Control of Bottom Pres-
box establishes a client-server connection be-
sure and Flow Rate. The cascade control
tween the Matlab environment and The Olga
scheme used here is shown schematically in
2000 through Olga Server so that the simulation
Figure 13. The master, outer, controller is a
results from Olga 2000 becomes available for
pressure controller while the slave one, inner, is a
Matlab on-line. This toolbox allows for advanced
flow controller. Both controllers are of standard
process simulation and control. In the following,
PI type. The controller parameters are shown in
all the control schemes are implemented in the
Table 5. The simulation starts with fully open
Matlab environment and the Olga model is con-
valve. The slug period is 200[sec]. The controller
trolled and steered from a Matlab application.
is switched on 350 seconds after start. The slug is

7
suppressed immediately and the setpoint is over. The responses are shown in Figure 17 and

reached after 4 minutes. The pressures, flow rate Figure 18. It is clear from Figure 18 that, with

and the valve opening are given in Figure 14 and the controller being active, we have been able to

Figure 15. achieve a lower top pressure, which in turn re-

5.3. Cascade Control of Top Pressure sults to higher production rate.

and Flow Rate. A disadvantage of former 6. Conclusions

control schemes is the need for underwater A dynamic Olga multiphase simulation model of

measurement of pressure. This is a costly and the Tiller loop has been developed and verified

sometimes unreliable measurement. The control against test data. The model has captured the

scheme in this section removes this disadvantage physical mechanisms of the slugs generated in

by using the pressure upstream the choke as con- the Tiller loop. Important phenomena such as

trol variable in the master loop. This control inverse response of the top pressure and asym-

structure is shown in Figure 16 metric step response of the bottom pressure have

been reported.

It is clear from simulation results that the top Several control strategies have been tested on the

pressure shows some sort of non-minimum phase developed model. It appears that a cascade con-

behaviour. This is a well-known fact that the trol strategy with feedback from bottom pressure

bandwidth of such a system is upper bounded. and flow rate at the top of the riser is best capa-

The inverse response is a constraint that cannot ble of suppressing the slugs. However, this con-

be removed with linear feedback control. The trol strategy requires pressure measurement at

controller must therefore have low proportional the bottom of the riser. In absence of such meas-

gain and high integration time. The issue of non- urement an alternative strategy might be imple-

minimum phase behaviour of top pressure has mented. This is described in Section 5.3. This

been discussed in more details in Section 0. control strategy has a weakness. Due to the non-

This controller was difficult to tune. Instead, as minimum-phase behaviour of the system, it suf-

often in practical cases in the fields, first the fers from a limited bandwidth and therefore slow

valve was chocked manually to 30% opening. control action. Simulation showed that the con-

After a short period of time the controller took troller was not capable of suppressing the slugs

8
by it self. The choke was first choked back and References
then the controller took over. After that the con- 1. Hollenberg, J.F. and Wolf, S. and
Meiring, W.J., “ A Method to Suppress
troller was perfectly capable of achieving and Severe Slugging in Flow Line Riser
Systems” , 7th BHR Group Ldt. et al
maintaining various setpoints. However, this is Multiphase Prod. Int. Cont., Cannes,
France, (1995).
not a major drawback. In practice, a well is stabi- 2. Havre, K. and Dalsmo M., “ Active
Feedback Control as the Solution to
lized manually about a higher bottom pressure. Severe Slugging” , SPE paper number
71540.
Then, the controller takes over and decreases the 3. Skofteland, G. and Godhavn, J-M., “
Suppression of slugs in Multiphase
bottom pressure gradually. Flow Lines by Active Use of Topside
Choke- Field Experience and Experi-
The controller with feedback from bottom pres- mental Results “, Proc. Of Multiphase,
San Romeo, Italy, (2003).
sure is preferable since the controller is faster 4. Havre, K. and Stray, H., “ Stabilization
of Terrain Induced Slug Flow in Multi
and reaches its pressure setpoint faster than a Phase Pipelines “, Servomøtet, Trond-
heim, Norway. (November 1999).
simple bottom pressure controller. 5. Courbot, A., “ Prevention of Severe
Slugging in the Dunbar 16" Multiphase
Pipeline” , Offshore Technology Con-
ference, Houston, Texas, (May 1996).
6. Godhavn, J.M., Fard, M. And Fuchs, P.
H., “ New Slug Control Strategies,
Tuning and Experimental Results “,
Journal of Process Control, Vol.15,
(August 2005).
7. Fard, Mehrdad P., “ The OLGA-
Matlab Toolbox”, Technical Report
No. NH-00036701, Norsk Hydro
ASA.(2001).

9
Figure 1 Diagram of the Tiller loop.

Figure 2 The topology of the pipelines

Figure 3 Frequency density and time response for the pressure at the top.

Figure 4 Frequency density and time response of the pressure at the bottom of the riser.

Figure 5 Comparison between the simulated and measured pressures at the bottom of the riser.

Figure 6 Comparison between the simulated and measured pressures at the top of the riser.

Figure 7 Comparison between the simulated and measured valve opening.

Figure 8 Inverse response at the top pressure with increasing valve opening.

Figure 9 Inverse response at the top pressure with decreasing valve opening.

Figure 10 Control structure with feedback from pressure at the bottom.

Figure 11 Simulated pressure response at the top and bottom of the riser with single pressure controller.

Figure 12 Simulated flow response and valve opening with single pressure controller.

Figure 13 . Cascade control structure of bottom pressure and volumetric flow.

Figure 14 Simulated pressure response at the top and bottom of the riser with cascade controller.

Figure 15 Simulated flow rate and valve opening with cascade controller.

Figure 16 Cascade control structure of top pressure and volumetric flow.

Figure 17 Simulation: Pressure responses with cascade control, feedback from top pressure and volumetric
flow rate.

Figure 18 Simulation: Flow rate and valve opening with feedback cascade control, feedback from top pres-
sure and volumetric flow rate.

10
Figure 1 Diagram of the Tiller loop.

11
Figure 2 The topology of the pipelines

12
Figure 3 Frequency density and time response for the pressure at the top.

13
Figure 4 Frequency density and time response of the pressure at the bottom of the riser.

14
Figure 5 Comparison between the simulated and measured pressures at the bottom of the riser.

15
Figure 6 Comparison between the simulated and measured pressures at the top of the riser.

16
Figure 7 Comparison between the simulated and measured valve opening.

17
Figure 8 Inverse response at the top pressure with increasing valve opening.

18
Figure 9 Inverse response at the top pressure with decreasing valve opening.

19
Set point

PC

Gas outlet

Platform
choke
Inlet separator
Oil outlet

Water outlet
PB

Figure 10 Control structure with feedback from pressure at the bottom.

20
Figure 11 Simulated pressure response at the top and bottom of the riser with single pressure
controller.

21
Figure 12 Simulated flow response and valve opening with single pressure controller.

22
Figure 13 . Cascade control structure of bottom pressure and volumetric flow.

23
Figure 14 Simulated pressure response at the top and bottom of the riser with cascade controller.

24
Figure 15 Simulated flow rate and valve opening with cascade controller.

25
Figure 16 Cascade control structure of top pressure and volumetric flow.

26
Figure 17 Simulation: Pressure responses with cascade control, feedback from top pressure and
volumetric flow rate.

27
Figure 18 Simulation: Flow rate and valve opening with feedback cascade control, feedback from
top pressure and volumetric flow rate.

28
X 0 100 110 190 215 215 215 217
Y 0 -0.17 -0.17 -1.57 -2 12.3 12.85 12.85
Table 1 X-Y coordinates of the pipe topology.

Valve opening Valve opening


Time Interval [sec]
Simulation Experiment
0-530 100% 100%
531-1500 Controller (46%) Controller (24%)
1501-2000 46% 24%
2001-3400 39% 16%
3401-5400 35% 12%
Table 2 Time intervals and corresponding mean value of valve opening.

Type Controller Setpoint KP Ti[min]


Simulation Master 1.9 1.5 1.35
Slave -10 0.1
Experiment Master 1.9 150 1.35
Slave -1 0.1
Table 3 Controller parameters for comparison study.

Setpoint [bar] KP Ti[Min]


2 2 2.5
Table 4 Pressure loop controller parameters.

Controller Setpoint KP Ti[Min]


Master 1.9 bar 1.5 1.3
Slave -10 0.10
Table 5 Controller parameter for cascade control.

29
Name : Mehrdad P. Fard
Company: Hydro

Work history:
1996-1997: Teaching Assistant, NTNU, Trondheim, Norway
2000- Now : Hydro

Education:
1996: MSc, Engineering Cybernetics, NTH, Trondheim, Norway
2001: PhD, Engineering Cybernetics, NTNU, Trondheim, Norway

Name : John-Morten Godhavn


Company: Statoil ASA

Work history:
1997-2001: Kongsberg Seatex ASA
2001- Now : Statoil ASA

Education:
1992: MSc, Engineering Cybernetics, NTH, Trondheim, Norway
1997: PhD, Engineering Cybernetics, NTNU, Trondheim, Norway

Name : Svein Ivar Sagatun


Company: Hydro

Work history:
1992-1998: ABB and TTS
1998- Now : Hydro

Education:
1988: MSc, Naval Architecture, NTH, Trondheim, Norway
1992: PhD, Engineering Cybernetics, NTNU, Trondheim, Norway

30

View publication stats

You might also like