Sei sulla pagina 1di 10

Slug Control With Large Valve Openings

To Maximize Oil Production


A.I. Ogazi, Y. Cao, H. Yeung, and L. Lao, Cranfield University

Summary
Severe slugging in an offshore riser pipeline imposes a major challenge to production and flow assurance in the oil and gas industry.
Riser-top-valve choking has shown effectiveness in eliminating
severe slugging. However, most manual-choking and active-control techniques were tuned by trial and error, resulting in an operation at a smaller-than-necessary valve position for a stable-flow
condition. This imposes unnecessarily high backpressure on the
riser pipeline, which leads to reduction in production. One way
to overcome this problem is to design the active-control system to
operate at a large valve position. However, at such an operating
point, the riser-pipeline system is naturally open-loop unstable
associated with severe slugging flows.
In this work, an approach to tune a robust proportional-integralderivative (PID) slug controller at an open-loop unstable condition
is proposed. First, at an open-loop unstable operating condition,
a reliable linear model is derived from the nonlinear simplified
riser/separator model (SRSM) developed in previous work. Then,
a robust stabilizing PID controller is designed on the basis of the
linearized model.
The controller was successfully applied to a 2-in. laboratory
riser at Cranfield University and an 8-in. generic industrial riser
system modeled in the commercial multiphase-flow simulator
OLGA. Simulation on the industrial riser system shows that the
proposed approaches not only can eliminate severe slugging but
also can increase oil production. It also shows that the percentage
improvement in oil production compared with manual choking will
increase as the well pressure declines. This means that adopting
active slug control is even more beneficial for mature oil fields
than for relatively new fields.
The result is very significant for mature fields that are susceptible to severe slugging and low oil production because of declining
reservoir pressure.
Introduction
Severe slugging in an offshore riser-pipeline system is one of the
most undesired flow regimes because of its potential to initiate and
sustain system instability. Because of the huge variation in pressure and flow associated with it, its consequences in oil and gas
production are a very serious concern. Severe pressure and flow
oscillations can cause deterioration of reservoir performance and
productivity, poor phase separation, compressor overloading, trips,
and production deferment. Conversely, a slug-control system will
eliminate or reduce the occurrence of these adverse conditions.
The primary objective of a slug-control system is to stabilize
the riser-pipeline system by suppressing severe slugging. Conventional solutions to address this severe-slugging issue include design
modification of upstream facilities (Fargharly 1987; Makogan
and Brook 2007), riser-base gas lift (Alvarez and Al-Malki 2003;
Cousins and Johal 2000; Duret and Tran 2002; Al-Kandari and
Koleshwar 1999; Jansen and Shoham 1994; Meng and Zhang
2001; Pots et al. 1987), gas reinjection (Tengesdal et al. 2003),
homogenizing the multiphase flow (Hassanein and Fairhurst 1998),
and installation of slug catchers and riser topside valve choking

Copyright 2010 Society of Petroleum Engineers


This paper (SPE 124883) was accepted for presentation at the Offshore Europe Oil &
Gas Conference & Exhibition held in Aberdeen, UK, 811 September 2009, and revised
for publication. Original manuscript received for review 5 June 2009. Revised manuscript
received for review 12 October 2009. Paper peer approved 16 November 2009.

812

(Schmidt et al. 1980). Among these solutions, choking transforms


the unstable flow in the riser to stable flow; however, it induces
extra backpressure on the pipeline. Active feedback, feed forward,
and cascade control systems have been applied to dynamic choking
for slug control (Henriot et al. 1999; Drengstig and Magndal 2001;
Jansen et al. 1996; Godhavn et al. 2005; Molyneux and Kinvig
2000; Storkaas et al. 2001; Storkaas and Skogestad 2004).
However, these methods have focused primarily on system
stability with little or no attention on their adverse effects on oil
production. Because of the difficulty of control tuning for openloop unstable systems, many slug controllers are designed to operate around an open-loop stable condition (i.e., the control valve
operating around a position where a manual choking method is able
to stabilize the system, usually at a smaller-than-necessary valve
position). This imposes unnecessarily high backpressure on the
riser-pipeline system and leads to a production reduction. In this
work, the application of an active feedback control tuned at wide
valve-opening positions, corresponding to an open-loop unstable
operating point, is studied. This technique achieves system stability
with minimum backpressure imposed on the riser pipeline, thereby
minimizing the effect of slug control on production.
In order to tune the controller at an open-loop unstable condition, a linear model is derived from a simplified riser model at the
unstable condition. Because the multiphase riser-pipeline system
is extremely nonlinear, to ensure the stability of the control system
for a wide operating range, the PID controller is designed on the
basis of a number of robust performance criteria. As a result, the
robustly designed PID controller is able to work at a much wider
valve-opening position, further reducing the controllers effect and,
under the right conditions, possibly even increasing production.
The paper begins with a description of the severe-slugging
phenomenon, methodology, and system description. Analysis is
then presented of the open-loop control of the laboratory 2-in.
riser system and the generic (industrial) riser system using bifurcation maps. This is followed by a description of the robust PID
tuning method for an open-loop unstable condition, together with
the implementation and effect on production of the controller
designed. Finally, analysis of stability and production in declining
reservoir-pressure condition based on the manual-choking and
active-control methods is presented before concluding the work.
Background
Severe Slugging. Severe slugging is a four-stage cyclic ow phenomenon. One major condition for severe slugging to occur is the
presence of a low point (dip) in the pipeline upstream of the riser.
This causes liquid to accumulate at the riser base because of the
balance of pressure by the opposing gravitational force, causing
the liquid to block the pipeline. With the pipeline blocked, gas
ow into the riser is stopped and further inlet gas is compressed
in the pipeline, resulting in pipeline pressure building up with a
continuous liquid accumulation in the riser. This continues until
the pipeline pressure overcomes the hydrostatic head in the riser
such that the liquid slug is pushed out of the riser. This results
in a pressure reduction in the pipeline, which allows the gas to
expand, penetrate the liquid, and increase the ow velocity. With
the gas front entering the riser, the liquid is blown out followed
by a drop in velocity and pressure. This causes the liquid to fall
back and block the riser base, resulting in a repeat of the cycle.
A detailed description of the severe-slugging phenomenon was
given by Taitel (1986).
September 2010 SPE Journal

TABLE 1FLOW RATE AND OPERATING PRESSURE CONDITIONS OF RISER SYSTEMS


Top Separator
Operating
Pressure (barg)

Fluid Source Type

Liquid (kg/s)

Gas (kg/s)

2-in. riser system


(experiment)

Constant flow rate

0.75

0.0033

8-in. generic system


(simulation)

30

Constant reservoir
pressure at 69 barg

Pressure
dependent

0.525

SRSM. To design a stabilizing controller for an open-loop unstable


system, it is necessary to have a reliable model of the system.
Extended from the simplied riser model (SRM) developed by
Storkaas et al. (2003), an improved SRSM has been developed
in previous work (Ogazi et al. 2009). Simulation results obtained
by using the improved SRSM closely match the experimental
results much better than those obtained by using the original SRM.
A detailed comparison between the SRM and the SRSM is reported
by Ogazi et al. (2009). Basic equations of the SRSM are provided
in Appendix A.
Experimental and Simulation Systems. In this study, experimental and simulation studies have been carried out for two riserpipeline systems. The rst system is a 2-in. riser-pipeline system
in the multiphase-ow laboratory at Craneld University. The
conguration and operating conditions of the system are described
in Appendix C. The second system is a generic (industrial) riser
system modeled using the commercial simulator OLGA, which
was developed by the SPT Group. The generic model is an 8-in.
riser-pipeline system consisting of a 5000-m-long pipeline, a 120m-high riser, and a pressure-driven well of 69 barg. An additional
gas lift source with a constant mass-ow rate of 0.525 kg/s is
applied at the wellhead.
Both systems are modeled using the SRSM in Matlab software.
Then, linear models required for controller design are obtained at
open-loop unstable points using the SRSM of both systems. The
controller designed for the 2-in. riser-pipeline system is implemented in the physical plant and tested through simulation using the
SRSM also; while the controller designed for the 8-in. generic riser
system is examined through OLGA simulation. The effect on oil
production is evaluated though OLGA simulation using the generic
model with a pressure-dependent well source. The flow-rate and
pressure conditions of both systems are summarized in Table 1.

Riser-base pressure, barg

Choking and Bifurcation Maps


To demonstrate the suitability of the SRSM for control design, a
series of manual-choking positions is tested on both riser systems.
The results predicted by the SRSM models are compared with
those obtained from experiment on the 2-in. riser system and
from OLGA simulation on the 8-in. generic model, respectively.
These results are presented in riser-base-pressure bifurcation maps
shown in Figs. 1 and 2 for the 2-in. and 8-in. riser systems,
respectively.

Riser-separator model
Experimental data

2.5
2
1.5
1

Bifurcation Map of the 2-in. System. Results from open-loop


simulation of the 2-in. riser-pipeline system using the SRSM (solid
line) and those through experiments (dashed line) are presented in
the bifurcation map shown in Fig. 1. The PRB oscillates between
minimum and maximum pressure points for large z (>30%). Both
the SRSM and the experimental results show that, at a critical z
of 25%, the desired stable nonoscillatory ow regime is obtained.
The corresponding PRB from simulation and from experiment are
2.25 and 2.27 barg, respectively.
Bifurcation Map of the 8-in. System. Results from open-loop
simulation of the 8-in. generic riser model using the SRSM with
well source (dashed line) and OLGA (solid line) are presented in
the bifurcation map shown in Fig. 2. For z between 100 and 13%,
the system is unstable and the PRB oscillates between minimum and
maximum pressure points. The desired stable nonoscillatory ow
regime is obtained at a critical z of 12% for both SRSM and the
OLGA simulation results. The corresponding PRB from the SRSM
and from OLGA simulation are 41.2 and 41.05 barg, respectively.
These critical values indicated by the bifurcation maps give
the minimum PRB and maximum z of the system to be stabilized
by manual choking. Our interest is to stabilize these systems on
unstable operating points, where the values of z are larger than
these critical values such that the total pressure drop across the
riser and the valve is reduced; thus, the overall production is
increased.
Robust PID Tuning Using Linear Models
Linear models of both systems are obtained using the SRSM at
open-loop unstable operating points. These linear models are then
used to obtain appropriate parameters for PID-controller tuning.
The PID-controller parameters are obtained to satisfy a number of
robust stability criteria. These robust stability criteria are described
in the next section with a brief introduction of the control-system
structure.
Control-System Structure and Sensitivity Functions. Fig. 3
shows the basic control structure applied to both riser-pipeline
systems for severe-slugging control as a feedback control system.
The manipulated variable u (controller output) is the valve opening,
z, and the controlled variable is the riser-base pressure, PRB.

Riser-base pressure, barg

System

Flow-Rate Condition

60

OLGA
SRSM with well source

55
50
45
40
35
30
0

10

20

30

40 50 60 70
Valve opening, %

80

90 100

Fig. 1Riser-base-pressure bifurcation map of the 2-in. riser.


September 2010 SPE Journal

10

20

30 40 50 60 70
Valve opening, z, %

80

90 100

Fig. 2Riser-base-pressure bifurcation map of the 8-in. generic


system.
813

d
Gd
PRB_setpoint

PRB_actual

n
Fig. 3Feedback control-loop diagram for severe-slug control.

In Fig. 3, n represents uncertainties caused by measurement


noise and modeling errors; Gd is the transfer function from disturbances to PRB, where disturbances d include the liquid- and
gas-flow-rate variations, well pressure, and downstream (topside
separator) pressure fluctuations; G is the transfer function of the
riser-pipeline system obtained through linearization of the SRSM;
and K is the PID-controller transfer function given as

1
K = Kc 1 +
+ D s , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (1)
is

where Kc is the controller gain, i is the controller integral time,


and D is the controller derivative time.
According to Fig. 3, the riser-base pressure of the riser-pipeline
control system can be represented as
PRB _ actual = Gu + Gd d , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (2)
where u, the controller output (valve opening, z), can be written as

u = K PRB _ setpoint PRB _ actual n . . . . . . . . . . . . . . . . . . . . . . . (3)


Substituting u in Eq. 3 into Eq. 2, we have

PRB _ actual = GK PRB _ setpoint PRB _ actual n + Gd d . . . . . . . . . . . (4)


Thus, we can obtain the closed-loop response of PRB_actual as
PRB _ actual = (1 + GK ) GKPRB _ setpoint + (1 + GK ) Gd d
1

(1 + GK ) GKn
= TPRB _ setpoint + SGd d Tn.
1

. . . . . . . . . . . . . . . . . . . . . . . . (5)
In Eq. 5, S = (1 + GK ) is known as the sensitivity function
1

and T = (1 + GK ) GK is known as the complementary sensitivity


function because S + T = 1. These sensitivity functions are used to
define the criteria for robust PID-controller design as follows.
1

Controller Design Criteria. The criteria used in determining suitable PID parameters are the lower and upper bounds on the loop
gain, GK; the upper bound on sensitivity function, ||S||; and the
lower bound on bandwidth, B*.
Upper Bound on Sensitivity Function Peak, ||S||. Good disturbance rejection, set-point tracking, and minimal noise transmission
are performance requirements that the controller aims to achieve to
ensure robust stability. Considering that a basic control objective is
to manipulate the controller output, u, so that PRB_actual is as close
to PRB_setpoint as possible, a set of requirements is imposed on Eq. 5.
814

From Eq. 5, the first requirement is rejection of disturbance effect,


d. These disturbances could be caused by liquid- and gas-flow-rate
variations or well-pressure and downstream (topside separator)
pressure fluctuations. The second requirement is the tracking of the
PRB set-point change. These two requirements are ideally achieved
with S 0 or T 1, which requires GK to be large. In reality, the
sensitivity function peak, ||S||, is typically required to be less than
2 (6 dB) at all frequencies where control is required. Larger values
of ||S|| indicate poor performance and poor robustness.
Upper and Lower Bound on the Loop Gain, GK. From Eq. 5,
the requirement for minimal noise transmission is achieved with
T 0 or S 1, which requires GK to be small. This places an
upper bound on GK required to satisfy T 1 and a lower bound on
GK required to satisfy T 0. Thus, there is a need for trade off in
the magnitude of GK to meet both requirements. This trade off is
provided in different frequency ranges in which these performance
requirements are most needed (Skogestad and Postlethwaite 2005).
Because noise is mainly of high frequency, most of these performance requirements can be achieved by designing the controller, K,
such that |GK| is large [i.e., |GK| > 1 at lower frequencies below the
crossover frequency (c)] or such that |GK| is small (i.e., |GK| < 1
at higher frequencies above the crossover frequencies). The crossover frequency is the frequency where |GK| first crosses 1 from
above (Skogestad and Postlethwaite 2005).
Lower Bound on Bandwidth, B*. Giving that the riser-pipeline
system has a pair of complex-conjugate unstable poles expressed
as p = x + yj, at the open-loop unstable operating point, the lower
bound on bandwidth given as

*B > 0.67 x + 4 x 2 + 3 y 2 . . . . . . . . . . . . . . . . . . . . . . . . . . (6)


is required for effective control, (Skogestad and Postlethwaite
2005). B* is the frequency at which |S| first crosses 0.707 (3 dB)
from below.
After G is obtained through linearization of the SRSM, the
controller parameters, Kc, i, and D, are designed to satisfy these
criteria.
Robust PID Tuning for the 2-in. System
According to Fig. 1, the 2-in. riser system is open-loop stable at z =
25% and unstable at z > 25%, with manual choking. Consequently,
we aim to stabilize the system at the open-loop unstable operating
point of z = 30%, using a PID controller. The transfer function of
the linear system model at this operating point is derived using
the model-linearization procedure described in Appendix B. The
transfer function of the linear model obtained at z = 30% valve
opening with z as input and PRB as output is obtained as

G30% =

54.11s 3 + 34270 s 2 + 28710 s 3.641 10 12


.
s + 632.5s 4 17.15s 3 + 245.3s 2 + 31.17s + 7.979 10 16
5

. . . . . . . . . . . . . . . . . . . . . . . . (7)
According to Eq. 7, we obtain a pair of complex-conjugate
poles, 0.0746 0.633i. From Eq. 6, the lower bound of B* is
September 2010 SPE Journal

TABLE 2PID TUNING PARAMETERS OBTAINED AT z = 30% FOR THE 2-in. RISER
Controller parameters

Stability parameters

Kc (barg )

|GK|
< c

||S||

0.001

800

0.005

5.6

0.0015

0.0014

158

2 10

0.05

800

0.005

14.06

2.19

6606

1.26 10

1.2

800

0.005

0.96

50.4

10.6

141253

2.37 10

obtained as 0.791 rad/s. ||S||, |GK|, and the system B* are evaluated by analyzing the Bode plot of |S| and |GK| at various controller parameters, Kc, i, and D. Table 2 shows the PID-controller
design parameters and corresponding ||S||, B*, and |GK| obtained
from the Bode plot.
From Table 2, parameters with Kc = 1.2 are the best at satisfying the stability criteria. With this set of PID-controller parameters,
the effectiveness of the controller is tested by implementing it
on the SRSM and on the flow-laboratory 2-in. riser system. The
experimental result obtained from the physical plant, as shown in
Fig. 4, indicates that the system is unstable (severe slugging) at
30% valve opening. However, when the controller is switched on
after 900 seconds, severe slugging is suppressed with PRB at 1.9
barg, corresponding to approximately 30% valve opening.
In simulation with the SRSM, a similar result is obtained
showing that, when the controller is switched on, severe slugging
is suppressed with PRB at 1.95 barg, corresponding to z at 27%.
Because of space limitation, this result is not included.

|GK|
> c

unstable operating of z = 20% using a PID controller tuned at this


point. The transfer function of the linear model obtained at z = 20%
valve opening with z as input and PRB as output is given as
G20% =

0.1037s 0.0002236
. . . . . . . . . . (8)
s 3 + 5.275s 2 0.001128s + 0.00001282

From Eq. 8, we obtain a pair of complex-conjugate poles,


0.0001 0.0016i. From Eq. 6, the lower bound of B* is obtained
as 0.00296 rad/s. ||S||, |GK|, and the actual B* are evaluated
by analyzing the Bode plot of |S| and |GK| at various controller
parameters, Kc, i, and D. Table 3 shows the PID-controller design
parameters and corresponding ||S||, B* and |GK| obtained from the
Bode diagram.
From Table 3, parameters with Kc = 2 are the best at satisfying
the stability criteria. Using this set of PID-controller parameters,
the effectiveness of the controller is tested by implementing it
on the generic riser model in OLGA and in the SRSM with well
source.
For the generic system, the simulation result shown in Fig. 5
indicates that, at fixed valve opening of 20% (open loop) the system is unstable; thus, the riser-base pressure, PRB, oscillates. When

Robust PID Tuning for the 8-in. System


From Fig. 2, the 8-in. generic riser system is open-loop stable at
z = 12% and open-loop unstable at z > 13% (with manual choking).
Consequently, we aim to stabilize the system at the open-loop

Riser- base pressure


100

80

1.5

60

40

0.5

20

0
1

1,001

2,001

3,001

4,001

5,001

Riser-top pressure
Valve opening, z

Pressure, barg

2.5

Valve opening

Time, s
Fig. 4Slugging-control experimental results on the 2-in. system.

TABLE 3PID TUNING PARAMETERS OBTAINED AT z = 20% FOR THE 8-in. RISER
Controller parameters

Stability parameters

Kc (barg )

||S||

|GK|
< c

|GK|
> c

0.05

500

0.005

14

0.0025

0.0020

158

5.7 10

0.1

500

0.005

2.6

0.0034

0.0241

316

1.1 10

500

0.005

0.0392

0.0353

6309

2.7 10

500

0.005

1.02

0.603

0.093

16788

2.4 10

10

500

0.005

1.03

0.813

0.187

30902

5.6 10

15

500

0.005

1.05

0.906

0.278

47836

7.9 10

September 2010 SPE Journal

815

46

60

42

40

38

20

34

50

10

12

14

16

18

30

ML-riser top

40

20

5.2

ML-well

30

4.4

20

3.6

10

2.8

10

12

14

16

18

Riser-base
pressure, barg

50

Valve opening
PRB

80

2
20

Well liquid
mass-flow rate, kg/s

Valve opening, %
Riser-top liquid
mass-flow rate, kg/s

50

Time, h
Fig. 5Slugging-control OLGA simulation results.

TABLE 48-in. SYSTEM STABLE OPERATION


COMPARISON
Active
Control

Manual
Choking

Variable
Valve opening, z (%)
Riser base pressure, PRB (barg)

12

20

57.6

41.05

36.8

34.4

the controller is switched on after 5 hours, with PRB set point at


36.8 barg, severe slugging is suppressed (system stabilized) with z
at 20%. By reducing the PRB set point further, the controller maintained system stability to a minimum riser-base pressure of 34.4
barg with valve opening increasing to a maximum value of 57.6%.
Beyond this point, the controller is not able to stabilize the system.
However, when the valve opening was returned to 20% fixed valve
opening (open loop) after 15 hours, the system gradually returned
to a totally unstable condition. These pressures and valve-opening
conditions are summarized in Table 4, in comparison with the
bifurcation-map results shown in Fig. 2.
The valve-opening (controller-output) trend in Fig. 5 appears to
be constant at the stable condition. However, this is not the case
because the valve opening is continuously being adjusted in small
steps by the controller to maintain stability. Fig. 6 shows the valveopening trend for the maximum stable valve opening.

Thus, the controller tuned at open-loop unstable operating point


of z = 20% stabilized the system at this same point and also is capable of stabilizing the system at a wider valve opening with the PRB
set point further reduced. The significance of these results is obvious
when the accumulated production at these pressure and valve-opening conditions is analyzed, as shown in the next sections.
Effect on Production
Production in a 24-hour simulation is analyzed to assess the effect
of open-loop-tuned controller compared with that in fixed choking
at the same valve positions. It must be mentioned that, when the
choke valve is fixed at these openings, the system is unstable with
large fluctuations in flow and pressure.
Fig. 7 shows that, from the OLGA simulation, the production
under severe-slugging condition at the valve position of z = 20% is
303.4 m3/d. By stabilizing the system with the controller designed
at z = 20%, as shown in the top part of Fig. 5, the production
obtained is 324.6 m3/d. With manual choking control, the system
is stabilized (severe slugging eliminated) at z = 12%, as shown in
Fig. 2, and the production is 280.25 m3/d, as shown in Fig. 8.
This implies that, by implementing the robust PID controller at
this operating point (z = 20%), production is increased by 7% when
compared to the production under severe-slugging condition at the
same fixed valve position and that production is increased by 15%
when compared to manual choking control at z = 12%.

57.62

Valve opening, %

57.619
57.618
57.617
57.616
57.615
57.614
57.613
13.7

13.75

13.8

13.85

13.9

13.95

14

Time, h
Fig. 6Valve-opening trend for slugging control, OLGA simulation results.
816

September 2010 SPE Journal

Liquid production, m3/d

350
340

Severe slugging
Active control

330
320
310
300
290
280
270
20

57.6
Valve opening, %

Fig. 7Production Comparison I, with and without active slug


control.

By implementing the controller at the maximum possible


operating point of z = 57.6%, the production obtained is 340
m3/d, as shown in Figs. 7 and 8. This implies that, by operating
the controller at this point, (z = 57.6%), production is increased
by 7.1% when compared to the production under severe-slugging
condition by fixing the valve at the same position and that production is increased by 21.3% when compared to manual choking
control at z = 12%.
Analysis of the production with the valve fully open (z = 100%)
shows that the production with severe slugging occurring at this
condition is 335 m3/d. This indicates that the production achieved
by the robust PID controller operating at z = 57.6% is even 1.4%
higher than that with the valve fully open. Similar results were
obtained through the SRSM simulation, but are not presented here
because of space limitations.
It can be observed that, for the flow and operating conditions
applied, higher-percentage increase in production is obtained with
comparison to manual choking, which reflects the high reduction
in production from the unstable condition if the manual-chokingcontrol method is implemented. Thus, manual choking control will
reduce production adversely. With these results, it is evident that,
with the robust PID tuning method, active slug control will meet
two fundamental objectives that previously have been considered
to be incompatible: stabilizing the system (at a larger valve position other than the manual-choking stable operating point) and
increasing oil production.

Liquid production, m3/d

Stability and Production in DecliningReservoir-Pressure Condition


As the well pressure declines (as fields mature), the differential
pressure between the topside pressure set point and the well source
decreases; thus, the fluid-flow rate is reduced. This would impose

further disturbance (instability) on the riser system such that further


action is required to stabilize the system. With topside valve choking
control, this further action would imply reducing the valve opening
further. Implementing manual choking or the robust PID controller
in this condition could have a serious effect on production.
Fig. 9 shows the minimum riser-base pressure and the maximum valve opening required to stabilize the generic (industrial)
riser model at declining reservoir pressure (from 69 to 45.3 barg)
by using manual choking and by using the open-loop unstable
tuned controller. To stabilize the system with manual choking, the
maximum valve opening is reduced from 12% at 69 barg to 7% at
45.3 barg. However, with the robust PID controller, the system is
stabilized at a wider maximum valve opening of 57% at 69 barg
and 42% at 45 barg.
This result shows that, with the manual choking, the minimum
pressure drop across the riser is much higher than the maximum
pressure drop across the riser obtained with the controller implementation for the range of the well pressures. The effect of this on
production is illustrated in Fig. 10.
From Fig. 10, it is observed that the percentage of increase
in production caused by the controller implementation increases
with declining well pressure. This reflects the significant reduction in production incurred because of the manual-choking-control
method as the well pressure decreases. Thus, with the robust PID
controller, significant proportional production increase is maintained with manual choking, while production is further reduced
with declining well pressure. This result also shows the degree
of robustness of the controller in tracking PRB set point and in
stabilizing the system with disturbances imposed by the declining
reservoir pressure and varying fluid-flow rate as required by the
stability criteria. It should be pointed out that this result is based on
a linear well model. If the reservoir model is nonlinear, the trend of
the effect of slug control on production may behave differently.
Conclusions
In view of the work reported in this paper, which focused on multiphase severe-slug control at open-loop unstable operating point
using robust PID controllers in an active feedback control loop,
the following conclusions can be drawn.
Active feedback control implemented at an open-loop unstable
operating point is effective in suppressing slug formation and
controlling severe slugging in multiphase-flow pipelines. Active
feedback control at open-loop unstable operating point can impose
the minimum possible backpressure on the riser-pipeline system,
lower than the high backpressure imposed by manual-choking
method and some existing active-control solutions designed around
manual-choking operating point. Because of the significant reduction in backpressure achieved by implementing severe-slug control
at open-loop unstable operating point with active feedback control,
oil production is increased from the system. With the robust PID
controller, the percentage of increase in production will increase
when compared to manual choking as the well pressure declines.

350
OLGA Model

330
310
290
270
No control
(z=100%)

Controller
(z=57.6%)

Controller
(z=20%)

Manual
choking
(z=12%)

Severe-slug control method


Fig. 8Production Comparison II, different stable and unstable operation modes.
September 2010 SPE Journal

817

45

Valve openingManual choking


Valve openingController
River-base pressureController
Riser-base pressureManual choking

80

42

60

39

40

36

20

33

Minimum riser-base
pressure, barg

Maximum valve opening, %

100

30

0
40

45

50
55
60
Well pressure, barg

65

70

Fig. 9System stability at declining well pressure.

Nomenclature
A = separator cross-sectional area, m2
= internal gas mass-ow area, m2
Ap = pipe cross-sectional area, m2
G = gravity, m/s2
h1 = liquid level upstream of the riser bend, m
hL = separator liquid height, m
HR = riser height, m
Hs = separator height, m
H1 = critical liquid level, m
K1 = valve coefcient
K2 = gas-ow coefcient
L3 = length of horizontal riser top section, m
mG1 = mass of gas in pipeline, kg
mG2 = mass of gas in the riser, kg
mL = mass of liquid in the riser, kg

mLin = separator-inlet liquid mass-ow rate, kg/s


mLout = separator-outlet liquid mass-ow rate, kg/s
mmix,out = total uid mass-ow rate at riser top, kg/s
MG = molecular weight of gas, kg/Kmol
MGin = mass-ow rate of gas into the system, kg/s
MLin = mass-ow rate of liquid into the system, kg/s
MLout = mass-ow rate of liquid out of the riser, kg/s
MGout = mass-ow rate of gas out of the riser, kg/s
MG1 = internal gas mass-ow rate, kg/s
Ps = separator-top pressure, barg
PRB = riser-base pressure, barg
PRT = riser-top pressure, barg
Qin = separator-inlet gas volume ow rate, m3/s
Qout = separator-outlet gas volume ow rate, m3/s
R = gas constant, 8.314 J/K/mol
T = system temperature, K
vG1 = internal gas velocity, m/s
VG1 = volume of gas in pipeline, m3
VT = total riser volume, m3
z = riser-top valve position
LT = average liquid volume fraction at riser top
G = gas volume fraction in the pipeline
LM = liquid mass fraction
G1 = pipeline gas density, kg/m3

Liquid production, m3/d

Liquid productionController
Liquid productionManual Choking
Increase in production due to controller

400

160

350

140

300

120

250

100

200

80

150

60

150

40

100

20

0
40

45

50

55

60

65

70

Increase in production because of


controller implementation, %

The open loop unstable tuned control is robust in changing flow


conditions and declining-well-pressure condition.
The total possible oil recovery from an oil field will be increased
by controlling severe slugging at open loop unstable operating
point using active feedback control, as shown by simulation on the
generic (industrial) riser OLGA model and the SRSM with a well
source. This result suggests that the operation life of a reservoir
might be extended by adopting a well-tuned slug controller.

Well pressure, barg


Fig. 10Percentage of production increase vs. well pressure.
818

September 2010 SPE Journal

G2 = riser-top gas density, kg/m3


L = liquid density, kg/m3
Acknowledgments
This work has been undertaken within the Joint Project on Transient Multiphase Flows and Flow Assurance. The authors wish
to acknowledge the contributions made to this project by the UK
Engineering and Physical Sciences Research Council and the following: Advantica, BP Exploration, CD-adapco, Chevron, ConocoPhillips, Eni, ExxonMobil, FEESA, Institut Franais du Ptrole,
Institutt for Energiteknikk, Petrleos de Venezuela Intevep, Petrobras, Petronas, Scandpower Petroleum Technology, Shell, SINTEF,
Statoil, and Total. The author wish to express their sincere gratitude
for this support. The authors wish to thank SPT Group for the use
of OLGA during this work and for providing a generic model of
a tie-back system for use in testing the slug-control algorithm.
OLGA is a registered trademark of SPT Group.
References
Al-Kandari, A.H. and Koleshwar, V.S. 1999. Overcoming Slugging Problems in a Long-Distance Multiphase Crude Pipeline. Paper SPE 56460
presented at the SPE Annual Technical Conference and Exhibition,
Houston, 36 October. doi: 10.2118/56460-MS.
Alvarez, C.J. and Al-Malki, S.S. 2003. Using Gas Injection for Reducing
Pressure Losses in Multiphase Pipelines. Paper SPE 84503 presented
at the SPE Annual Technical Conference and Exhibition, Denver, 58
October. doi: 10.2118/84503-MS.
Cousins, A.R. and Johal, K.S. 2000. Multi purpose riser. European Patent
No. EP1022429 (A1).
Drengstig, T. and Magndal, S. 2001. Slug control of production pipeline.
Technical Report N-4091, School of Science and Technology, Stavanger University College, Stavanger, Norway.
Duret, E. and Tran Q. 2002. Gas injection controlling liquid slugs in pipeline riser. European Patent No. GB2374161 (A).
Fargharly, M.A. 1987. Study of Severe Slugging in Real Offshore Pipeline
Riser-Pipe System. Paper SPE 15726 presented at the Middle East Oil
Show, Bahrain, 710 March. doi: 10.2118/15726-MS.
Godhavn, J.-M., Fard, M.P., and Fuchs, P.H. 2005. New slug control strategies, tuning rules and experimental results. Journal of Process Control
15 (5): 547557. doi: 10.1016/j.jprocont.2004.10.003.
Hassanein, T. and Fairhurst, P. 1998. Challenges in the mechanical and
hydraulic aspects of riser design for deep water developments. Presented at the 21st Offshore Pipeline Technology Conference, Oslo,
Norway, 2324 February.
Henriot, V., Courbot, A., Heintz, E., and Moyeux, L. 1999. Simulation of
Process to Control Severe Slugging: Application to the Dunbar Pipeline. Paper SPE 56461 presented at the SPE Annual Technical Conference and Exhibition, Houston, 36 October. doi: 10.2118/56461-MS.
Jansen, F.E. and Shoham, O. 1994. Methods for Eliminating Pipeline-Riser
Flow Instabilities. Paper SPE 27867 presented at the SPE Western
Regional Meeting, Long Beach, California, USA, 2325 March. doi:
10.2118/27867-MS.
Jansen, F.E., Shoham, O., and Taitel, Y. 1996. The elimination of severe
slugging experiments and modeling. Int. J. of Multiphase Flow 22
(6): 1055-1071. doi: 10.1016/0301-9322(96)00027-4.
Makogan, T.Y and Brook, G.J. 2007. Device for Controlling Slugging.
International Patent No. WO/2007/034142.
Meng, W. and Zhang, J.J. 2001. Modeling and Mitigation of Severe Riser
Slugging: A Case Study. Paper SPE 71564 presented at the SPE Annual
Technical Conference and Exhibition, New Orleans, 30 September
3 October. doi: 10.2118/71564-MS.
Molyneux P.D. and Kinvig, J.P. 2000. Method of eliminating severe slugging in a riser of a pipeline includes measuring pipeline pressure and
operating a valve. European Patent No. GB2358205 (A).
Ogazi, A.I., Ogunkolade, S.O., Cao, Y., Lao, L., and Yeung, H. 2009. Severe
slugging control through open loop unstable PID tuning to increase
oil production. Proc., 14th International Conference on Multiphase
Production Technology, Cannes, France, 1719 June.
Pots, B.F.M., Bromilow, I.G., and Konijn, M.J.W.F. 1987. Severe Slug Flow
in Offshore Flowline/Riser Systems. SPE Prod Eng 2 (4): 319324;
Trans., AIME, 283. SPE-13723-PA. doi: 10.2118/13723-PA.
September 2010 SPE Journal

Schmidt, Z., Brill, J.P., and Beggs, H.D. 1980. Experimental Study of
Severe Slugging in a Two-Phase-Flow PipelineRiser Pipe System.
SPE J. 20 (5): 407414. SPE-8306-PA. doi: 10.2118/8306-PA.
Skogestad, S. and Postlethwaite, I. 2005. Multivariable Feedback Control:
Analysis and Design, second edition. West Sussex, UK: John Wiley
& Sons.
Storkaas, E. and Skogestad, S. 2004. Cascade control of unstable systems
with application to stabilization of slug flow. Presented at IFAC Symposium ADCHEM 03, Hong Kong, January 2004.
Storkaas, E., Skogestad, S., Alstad, V., Undeli, M., and Havre, K. 2001.
Stabilisation of desired flow regimes in pipelines. Paper 287d presented
at the AIChE Annual Meeting, Reno, Nevada, USA, 9 November.
Storkaas, E., Skogestad, S., and Godhavn, J.-M. 2003. A Low Dimensional
Model of Severe Slugging for Controller Design and Analysis. Presented at Multiphase 03, San Remo, Italy, 1113 June.
Taitel, Y. 1986. Stability of severe slugging. Int. J. Multiphase Flow 12 (2):
203217. doi: 10.1016/0301-9322(86)90026-1.
Tengesdal, J.., Sarica, C., and Thompson, L. 2003. Severe Slugging
Attenuation for Deepwater Multiphase Pipeline and Riser Systems. SPE
Prod & Fac 18 (4): 269279. SPE-87089-PA. doi: 10.2118/87089-PA.

Appendix ASRSM Equations


Conservation Equations. The SRSM consists of ve dynamical
state equations that account for mass of gas in the pipeline, mG1,
mass of gas in the riser, mG2, mass of liquid in the riser, mL, separator top pressure, Ps, and separator liquid height, hL, which are
dmG1
= MGin MG1 , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-1)
dt
dmG 2
= MG1 MGout , . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-2)
dt

dmL
= M Lin M Lout , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-3)
dt
dPs
Ps

dh
=
Qin Qout ) + A l , . . . . . . . . . . . (A-4)
(

dt
dt
A ( H s hl )
and
dhl ( mLin mLout )
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-5)
=
dt
A L
State-Dependent Variables. The state-dependent variables, such
as the riser-base pressure, PRB, and riser-top pressure, PRT, are
calculated using the ideal-gas law given as
PRB =

mG1 RT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-6)
VG1 MG

and
PRT =

mG 2 RT
, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-7)
VG 2 MG

where the volume of gas in the pipeline, VG1, and in the riser, VG2,
is calculated as
VG1 = G AP L . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-8)
and
VG 2 = VT VLR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-9)
819

G =

mGin

G
PRB
mGin mLin
+
G
L

. . . . . . . . . . . . . . . . . . . . . . . . . . (A-10)

VT = AP ( H R + L3 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-11)

variable, and y = output variable. The matrices in the linearized


model are obtained as partial differentials of the dynamical state
equations, given as
A=

f
x , u , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-5)
x

B=

f
x , u , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-6)
u

C=

g
x , u , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-7)
x

D=

g
x , u , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-8)
u

The internal gas mass-flow rate (riser base), MG1 is given as


MG1 = vG1G1 A , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-12)
where

G1 =

mG1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-13)
VG1

and
vG1 = ( h1 < H1 ) K 2

H1 h1
P PRT L g L H R
RB
.
G1
h1

f1
f1
x $ x
1
5

f
= % ' % , . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-9)
x
f
f5

5 $
x5
x1

. . . . . . . . . . . . . . . . . . . . . (A-14)
Total fluid flow out of the riser, mmix,out, is calculated using a
simplified valve equation given as
mmix,out = K1z

T ( PRT Ps ) . . . . . . . . . . . . . . . . . . . . . . (A-15)
g

T = LT L + (1 + LT ) G 2 . . . . . . . . . . . . . . . . . . . . . . . . (A-16)
G 2 =

mG 2
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-17)
VG 2

The gas and liquid mass-flow rate out of the riser is therefore
calculated as

MGout = 1 Lm mmix,out . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-18)


and

( )

M Lout = Lm mmix,out . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-19)

Appendix BLinearizing the Model


For a nonlinear system, the general state space form is given as
x = f ( x , u ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-1)
and
y = g ( x , u ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-2)
The nonlinear Eqs. B-1 and B-2 can be linearized about the steady
state ( x , u ) to obtain state equations of system which are of the
form
x = Ax + Bu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-3)
and
y = Cx + Du , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-4)

where A = state matrix, B = input matrix, C = output matrix, D =


input/output direct coupling matrix, x = state variable, u = input
820

f1
u
f , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-10)
= %
u
f5

u
g g
g , . . . . . . . . . . . . . . . . . . . . . . . . . . (B-11)
$
=
x x1
x5
and
g
= 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-12)
u
The state variables, x, are the dynamical state equations stated
in Appendix A. We can observe that the model presented in Appendix A is in the form of a differential algebraic equation. To obtain
the transfer function, the model is transformed from the differential
algebraic equation to an ordinary-differential equation. From Eq.
B-3, we can obtain the transfer function for the state variables as
shown in Eq. B-13 through Laplace transformation.
x( s )
u( s )

1
= ( Is A) B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-13)

And by substituting the Laplace transform of Eq. B-4 into Eq. B-13,
we obtain the transfer function for the output variable, PRB, as
y( s )
u( s )

= C ( Is A) B , . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-14)
1

where I is a (nn) unity matrix (n = number of states).


Appendix CMultiphase Facility at
Cranfield University
The facility consists of a 2-in. and a 4-in. riser-pipeline system
that can run alternatively. The 2-in. riser is a vertical riser with
upstream pipeline length of 39 m inclined downward at 2 and a
riser height of 11 m, while the 4-in. riser is catenary, with upstream
pipeline length of 55 m, also inclined downward at 2, and a riser
height of 10.5 m. Fluid of both systems is supplied from three
September 2010 SPE Journal

Fig. C-1Multiphase facility at Cranfield University.

independent single-phase sources for oil (dielectric 250), water,


and air. For each riser system, the supplied fluid mixes at a mixing point into the pipeline, which connects to the riser. The top of
both risers is equipped with a topside processing facility, which
includes a control valve and a two-phase vertical separator that
separates the fluid into liquid and gas for measuring instruments.
The two-phase separator is approximately 1.2 m high and 0.5 m
in diameter. It consists of the gas- and liquid-outlet control valves
and pressure, flow, temperature, and level transmitters. Pressure
and flow measurements are obtained at riser inlet and outlet.
A schematic of this facility is shown in Fig. C-1.

Anayo Isaac Ogazi holds a BEng degree in electrical/electronic engineering from Nnamdi Azikiwe University, Awka,
Nigeria, and an MS degree in process and system engineering
from Cranfield University. Currently, he is a PhD research student
in the Department of Offshore, Process and Energy Engineering
at Cranfield University, with research interest in robust control
systems for unstable offshore pipeline production systems. Yi
Cao holds an MS degree in control engineering from Zhejiang

September 2010 SPE Journal

University, China, and a PhD degree in engineering from the


University of Exeter. He is a senior lecturer with the School of
Engineering, Cranfield University. His research interests are in
advanced process control, including plantwide process control, nonlinear system identification, nonlinear model predictive
control, and process monitoring. Hoi Yeung holds a BS degree
in mechanical engineering from the University of Hong Kong
and a PhD degree from the University of Newcastle upon Tyne.
He is a reader and the head of Process Systems Engineering,
Cranfield University. His research interests are fluid engineering,
multiphase flows, and oil and gas production technology. Liyun
Lao is a research fellow in the Process Systems Engineering
group, Cranfield University, and a corporate member of the
Institute of Physics. He holds BS and MS degrees in instrumentation sciences and engineering from Zhejiang University. Lao
holds a PhD degree from the Department of Control Sciences
and Engineering, Zhejiang University, China. Before joining
Cranfield University, he worked in the area of two-phase flows
in vertical pipes for six years at Imperial College and University
of Cambridge. His research interests are experimental and
analytical investigation of gas/liquid flows in vertical pipes, multiphase-flow measurements, and process intensification relating to gas/liquid flows.

821

Potrebbero piacerti anche