SPE 56832
Automatic Control of Unstable Gas Lifted Wells
Bård Jansen, ABB Industri AS; Morten Dalsmo, ABB Corporate Research; Lars Nøkleberg, SPE, ABB Corporate
Research; Kjetil Havre, ABB Corporate Research; Veslemøy Kristiansen, ABB Industri AS; Pierre Lemetayer, SPE, Elf
Exploration Production.
Copyright 1999, Society of Petroleum Engineers Inc.
This paper was prepared for presentation at the 1999 SPE Annual Technical Conference and
Exhibition held in Houston, Texas, 3–6 October 1999.
This paper was selected for presentation by an SPE Program Committee following review of
information contained in an abstract submitted by the author(s). Contents of the paper, as
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
for commercial purposes without the written consent of the Society of Petroleum Engineers is
prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300
words; illustrations may not be copied. The abstract must contain conspicuous
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.
Abstract
Oil production wells on gas lift are sometimes unstable at low
gas lift rates, even though steady-state flow analysis gives
most efficient production at these gas lift rates. Unstable
production, often called heading, may lead to periods of
reduced or even no liquid production followed by large peaks
of liquid and gas. This results in average oil production less
than expected, and oil and gas production less than the
systems design capacity to allow for the production peaks
without causing shutdowns.
To solve the problem the amount of lift gas is normally
increased beyond the most optimum rate. When the lift gas
supply is limited, other gas lifted wells must then be shut in.
This paper describes one field proven sequence based
automation for oil production, plus a new model-based
automatic controller. Both technologies solve the problem of
unstable production from gas lifted wells through
manipulation of the production and/or the gas-injection
chokes. The sequence-based automation and the model-based
controller stabilize the production at operating points that
would be unstable under standard operation. Field
measurements such as wellhead pressure, annulus pressure etc.
are inputs to the controllers (or are used by the controller).
Examples are given of the model-based controller working
together with realistic transient flow models of gas lift wells.
Testing of the model-based controller on real unstable wells is
planned.
Introduction
In production systems with gas lifted and naturally flowing
wells, the problem of lift-gas allocation is well known. Many
companies still use the traditional gradient method in order to
allocate the lift gas between wells optimally so that the total
average oil production is optimized when the available amount
of lift gas is limited. However, this approach has several
disadvantages. First it does not account for the fact that the
wellhead pressures and flow rates are mutually dependent due
to the pressure drop in the gathering system. Secondly, and
sometimes more important, it does not account for heading
problems. Last but not least, the problem of proper and safe
unloading is also not addressed. Thus, an automation system
that handles oil production should take all these factors into
account.
Currently, there is an increasing interest within the oil
industry for addressing oil production optimization using
control system technologies and optimization techniques.
There are several reasons for this:
1. Reduced oil prize gives an increased focus on costs and
production efficiency.
2. Many reservoirs are today depleting. This means that the
total oil production is dependent of the well performance
rather than capacity limitations of the processing plant
3. Taxes and pressure on energy utilization has increased in
a number of areas recently. For example, the Norwegian
CO2-tax for the offshore industry is 50USD/metric ton
CO2 produced. This gives an enormous incentive to
reduce energy consumption on the installations.
Looking at other industries, such as Ethylene cracking and
polymerization, the possibility for increasing throughput
through a systematic approach in order to avoid downtime, de-
bottleneck the process and continuously monitor the
performance is in these industries crucial to their profitability.
It is obvious that the oil industry has much to learn regarding
these matters.
Possible benefits by increasing the level of automation in
the oil industry are:
1. Increased safety due to smooth behavior during
continuos production and start up, and shut in of wells
2. Increased production due to less downtime, faster
startup and continuously optimized production
3. More efficient use of manpower through
simplification of the well handling, and stable
behavior of the wells
2 B. Jansen, M. Dalsmo, L. Nøkleberg; K. Havre, V. Kristiansen, P. Lemetayer
Statement of the problem
Typical gas lifted wells have a stable behavior at elevated gas
injection rates and unstable behavior at low gas injection rates.
This means that a gas lifted well is not producing the
maximum possible amount of oil at low gas injection rates in
spite of the fact that these wells are operated most efficiently
at these injection rates. Unstable operational conditions are the
most important reason for this.
Operating a gas lifted well under unstable conditions has
several disadvantages. First, the full lift potential in the gas is
not properly used, resulting in a very inefficient operation.
Second, surges in the production facilities may be so huge that
severe operational conditions are likely to occur. Third,
production control and allocation becomes very difficult.
Stability Problems
Unstable operational conditions may occur in a gas lift well
because the characteristics of the system are such that small
perturbations can degenerate into huge oscillations in the flow
parameters. Unstable production, often called heading, may
lead to periods of reduced or even no liquid production. To
understand why heading occurs consider Fig. 1 that illustrates
the stability region for a typical lift gas performance curve. A
similar curve was first presented in Ref. 1. A typical gas lifted
well configuration is shown in Fig. 2.
At the highest gas injection rates, the pressure drop in the
tubing is dominated by friction. If the GOR (Gas Oil Ratio)
rises, the tubing pressure will increase which will reduce the
gas injection rate. This region therefore ensures stable
production and explains why well stabilization by increased
gas injection can be successful.
At low gas injection rates however, the hydrostatic
pressure gradient dominates the pressure drop in the tubing. A
small increase in GOR results then in a lower tubing pressure,
which leads to a higher gas injection rate from the annulus into
the tubing through the down-hole gas lift valve. Since the gas
rate is restricted by a gas injection choke at wellhead, the gas
pressure in the annulus will be reduced. After a time the gas
rate into the production tubing will therefor be reduced, with
resulting lower oil production rates. At low gas injection rates
the well is therefore intrinsically unstable in spite of the fact
that wells are operated most efficiently on the upward slope of
the LPR curve; cf. Ref. 1.
To further illustrate the stability problem, a stepwise
description of a heading cycle is given below:
1. Starting with an annulus pressure down-hole that is lower
than the bottom-hole pressure, there is no gas flow
through the down-hole gas lift valve into the tubing.
Production rate and gas/liquid ratio is low. Gas is injected
through the gas injection choke and annulus pressure
builds up.
2. After some time, the annulus pressure exceeds bottom-
hole pressure, and gas is injected into the tubing through
the down-hole gas lift valve.
3. The injected gas lightens the tubing gradient so that
bottom-hole pressure begins to decrease. Simultaneously,
SPE 56832
the production rate and wellhead tubing pressure begin to
increase.
4. Gas now flows from the annulus in to the tubing at an
increasing rate. Because insufficient gas can be supplied
through the gas injection choke, annulus pressure
decrease rapidly.
5. Oil and gas are produced through the production choke at
a high rate. Wellhead tubing pressure passes through a
maximum and bottom-hole pressure passes through a
minimum.
6. With decreasing annulus pressure, gas flow through
down-hole gas-lift valve decreases. The gradient in the
tubing becomes heavier and bottom-hole pressure
increases. The production rate and wellhead tubing
pressure decreases again.
7. When bottom-hole pressure exceeds annulus pressure, gas
injection into the tubing stops. With continued gas
injection rate at the wellhead, annulus pressure starts to
build again.
Typical unstable gas lift well production is shown in Fig. 3
and Fig. 4 for a typical North Sea well at two different gas
injection rates. The simulations were generated using the
dynamic multi-phase flow simulator OLGA (see e.g. Ref. 2),
which also was used to calculate the resulting lift performance
relationship curves shown in Fig. 5. Here the average oil
production rates are shown for different gas lift rates. As seen,
the loss in average production is high for unstable production.
Unstable production of gas lifted wells cause many
drawbacks, surge is not in agreement with smooth operation
and it implies safety aspects and shutdown risks. The total oil
and gas production must usually be less than the systems
design capacity to allow for the peak production. Unstable
mode often decreases sharply the lift gas efficiency. In
addition, difficulties with gas lift allocation computation due
to instabilities are also common. Well instabilities also induce
other drawbacks on facilities and well operations on
equipment.
Since heading can be caused by a large variety of factors,
such as incorrect gas lift string design, improper valve setting,
wrongly sized injection valve port, variation in supply
pressure, or valve leaking or plugging, it is often difficult to
find the origin of the heading. As a result, a pragmatic
approach is often used to solve the problem in short term. For
example, if a well is heading the operator often increases the
amount of lift gas or increases the back-pressure by adjusting
the wellhead production choke to a smaller opening (choking).
Although these methods can be effective in reducing heading,
production is still sub-optimal as either too much lift gas is
used (high cost and limited availability of lift gas) or the well
is produced against a high back-pressure (at low rate). In most
cases, too much gas is injected into the gas lifted wells or the
production rate is not maximized.
Reducing the size of the down-hole gas lift valve orifice
may also reduce or solve the heading problem, but the
resulting higher annulus pressure and gas injection pressure
lead to higher costs.
SPE 56832 AUTOMATIC CONTROL OF UNSTABLE GAS LIFTED WELLS
Proven technology
In 1990 Elf Exploration Production developed a system for
automatic handling of single wells and complete field sections.
Some of the results were reported in Ref. 3 and in Ref. 4. The
name for these technologies are “Full control of wells” or
FCW. FCW consists of two main parts:
1. An individual well management level, referred to here as
Monowell
2. A collective well management level referred to here as
Multiwell, which controls a series of wells in relation to
the oil and gas facilities.
The system is the first and still the unique field-applied
dynamic control that uses both the oil production choke and
the lift gas control valve of each well under continuous gas-
lift, see Fig. 6. The control schemes account continuously for
simple surface measurements on the wells and on the facilities.
Control concepts
Well production is optimized on the basis of technical
constraints and economical and strategic objectives, while
taking into account safety and production rules, reservoir
extraction policy, (max. flow rate per well, production quotas
etc.), well bore-formation interface (sand control, max. dP
etc.) and capacity of the installations (upstream &
downstream). The system is based on universal sequences and
fuzzy logic without any computation. It is a sequence-based
control dynamically adjusting lift gas and oil choke.
The technology features individual well management
including effective but gentle and controlled unloading of the
well that overcomes the energy threshold of the well. The
system is able to automatically reduce the injection gas flow
rate until a desired level or the instability region, and
thereafter continuously adjust the gas flow to assure stability
of the flow and pressure in the well. The control stabilizes
operating points that are unstable under standard operation
through an enhanced path during unloading and transition. The
instability is field proven to be path dependent.
In case of substantial variations in flow rate, the
adjustments on the well are modified immediately to bring the
well back to its normal regime.
The Multiwell control module ensures collective
management of wells by sequential start and restart of a group
of wells, for example all wells at an offshore platform. The
wells are started up in pre-determined sequence. Optimization
of the gas lift network is based on selective load shedding of
the wells. Based upon measurements e.g. separator level,
export pressure or lift gas supply pressure, the system
automatically observes bottlenecks that might lead to process
shutdown. In such cases the system automatically either
reduces the flow from the wells in a pre-specified order or
shutting in a number of wells.
3
Fig. 7 gives a real example from Grondin field that shows
how FCW can influence the behavior of a well; more
production with less lift gas.
General Architecture
The system includes the following major elements (see
Fig. 8): well equipment (with gauges and actuators), gas lift
and oil networks (with gauges), a local control system, and
one or more operator stations.
The well equipment required for each well includes an
actuated production choke and a gas lift injection control valve
plus transmitters to measure wellhead and casing pressure,
injection gas flow rate and optionally liquid flow rate,
production choke position and oil production choke pressure
drop.
The casing and tubing valves used for shut down are
managed only by the PSD and ESD systems. Only information
about shut down is given by the PSD and ESD systems to the
automation system. Fig. 9 shows a topology drawing of the
instrumentation.
The local control system commands the wells, receives
measurements from the various sensors installed in the wells
and in the gas and oil networks, controls gas injection flow
rate and liquid production. Program logic is implemented here.
The operator station is used in order to monitor the various
measurements, perform commands like well start, well stop
and selection of operating mode, consult the well history,
diagnose system operation and update parameters.
Experience
Today these technologies have been used for nearly ten
years on more than 200 wells. The experience shows that the
described concepts for automation increase the average oil
production from 5 to 20%, and a decrease in gas lift usage
from 5 to 20%.
In this article we would like to elaborate how these
benefits occurs.
Increased Process Uptime
Automatic control of the wells means both smooth and
efficient start-up. Thus, the start-up surge is dampened, and
the well is stable after the start-up phase. Especially there is no
more heading production mode. The major need of operation
for the oil production chokes and the lift gas injection control
valves is during the restart-up phase. In case of instability, the
control program achieves quickly corrective actions including
a choke back which dampers the surges.
The total production throughput is often limited by the
capacity in the separators. This means that the separator train
is heavily loaded. A fluctuation in flow or composition from
the wells leads to disturbances in the separation process. This
gives sometimes high-high alarms of level or pressure in the
separator and thereby process shut down. The control system
counteracts efficiently on up-coming problems downstream.
4 B. Jansen, M. Dalsmo, L. Nøkleberg; K. Havre, V. Kristiansen, P. Lemetayer
Increased Life of the Wells and Fewer Well works
Steady production also means steady bottom-hole pressure.
This is crucial for prevention of formation damage and
troubles on the draw down zone; i.e. the risk for collapse is
reduced (see Ref. 3).
An unstable production mode causes numerous and quick
pressure changes which increase the formation damage.
Equally important, the kick-off of a gas lifted well leads to
rapid and big changes of pressure between annulus and tubing
due to the inflow of gas in annulus, followed by opening of
unloading valves and gas lift valve. If not properly controlled,
these variations might lead to damages or even collapse of the
tubing. Formation damage usually decreases the reservoir
inflow performance (PI).
In such a situation workover operations in order to repair
the damages will lead to downtime and costly rig operations.
Using automation, the risk for damages will be reduced due to
better control and less pressure fluctuations.
Increased Production Rates from the Wells
The increase of average production rates is due to both
stable flow and shorter periods with reduced flow or shut
down. The stable flow increases the lift gas efficiency
especially under low flow rate of lift gas and increases the
usable capacity of the facilities.
Improved Drainage of the Reservoir
For some wells, water coning decides the water cut, such
that higher coning gives higher water cut. The coning is often
a function of the dynamics when the well is kicked off.
Smooth behavior of the well (also during kick off) gives
reduced coning and thus lower water cut.
Reduced use of lift gas for gas lifted wells
Economics and lift gas availability often require to operate
wells under a reduced lift gas injection, for example when
watercut increases or on strong capacity wells. Conversely,
high lift gas injection into tubing is usually required to
properly achieve the unloading during the restart phase. A
temporary high lift gas flow rate (see Fig. 7) in conjunction
with the oil choke operation achieves a successful restart.
Furthermore, it is often essential to reduce lift gas usage
during a well restart phase. This is achieved mainly due to
shortened start-up time, but also due to sequenced start-up and
the annulus packing. During continuous operation, a stable
flow requires lower lift gas support. Also, the sustained stable
conditions in the annulus gives less fluctuations in the lift gas
rate.
A New Model-Based Automatic Control Approach
In this section we will present a new model-based automatic
control approach for gas lifted wells. This method heavily
relies on dynamic (transient) gas lifted well models, and can
partly be viewed as further development of the method
presented in the previous section. It is another approach for the
automatic stabilization of gas lifted oil wells and thus
SPE 56832
maximizing the lift gas efficiency, which captures the well and
process knowledge in a dynamic model.
The main ingredient in the model-based gas lifted well
controller concept is a dynamic (transient) model of a gas
lifted well from which a model-based stabilizing gas lift
controller can be designed. By using such a model-based
concept it is possible to stabilize the pressures, temperatures
and flow rates of a gas lifted well in an operating point that is
unstable in open-loop (i.e., when no active control is used).
The model-based stabilizing gas lifted well controller makes
sure that the control error, the difference between the
(externally given) optimal reference operating point and the
real operating point, at any time is kept at a minimum. An
appealing feature of the model-based stabilizing gas lifted well
controller concept is that it is able to stabilize gas lifted wells
with different measurement devices (sensors) available for
control purposes.
Fig. 10 shows a schematic of the new model-based
stabilizing gas lift controller structure. The model-based gas
lift controller uses an externally given optimal reference point
and one or more process measurements (or a model-based
estimate of these) to calculate the opening of the production
choke and/or the gas injection choke. The preferred mode for
the externally given optimal reference point is the
specification of the optimal LGR (gas rate through the down-
hole gas injection valve).
Consider again Fig. 5 that shows a lift performance
relationship curve for a North Sea gas lifted well generated
using the dynamic multi-phase flow simulator OLGA. The
shape of the lift performance curve is typical for gas lifted
wells. The solid curve corresponds to injecting lift gas directly
through the lowest gas injection valve at constant rate.
Obviously, this is not possible, and the dotted curve shows the
resulting average production when the lift gas is injected at
constant rate through the gas injection choke. In both cases
the resulting oil production rate are shown as a function of the
amount of injected lift gas. Due to unstable production at low
gas injection rates when injection the lift gas through the gas
injection choke, it is seen from Fig. 5 that the typical loss in
average production is high for unstable production conditions
(dotted curve) as compared to stable operating conditions
(solid curve). The results in Fig. 5 agree very well with what
was reported in Ref. 1 where a transient simulator was used to
remedy gas lift problems. However, a transient simulator
alone may only help you to figure out how to improve your
gas lift design, whereas capturing the dynamic model in an
active control algorithm solves the stability problems without
changing the design. Indeed, experiments have shown that
with the new model-based stabilizing active controller the well
in Fig. 5 may be operated on or in infinitesimal distance from
the stable solid curve.
In Fig. 5, it is seen that there is no production when the lift
gas injection is zero. Note that this figure is not universally
valid. In fact:
1) For high PI wells, the solid curve (stable flow) gives
production with no injection but the dotted line (unstable)
gives no production for no injection.
SPE 56832 AUTOMATIC CONTROL OF UNSTABLE GAS LIFTED WELLS
2) For low PI wells, the dotted curve has also no production
for no injection but the solid curve crosses the dotted
curve. Below a minimum lift gas, no stable flow is
possible whatever the control is. Below the cross point, it
is more economical to produce unstable (intermittent gas -
lift).
A Simple Dynamic Model Structure
The idea behind the new model-based concept is to analyze
and design stabilizing controllers, and, if applicable,
estimators based on a dynamic model of the system. For this
purpose we have developed a structure for a simplified
dynamic non-linear model based on physical principles of gas
lifted wells suitable for controller and estimator design. The
main purpose with this dynamic model is to describe the
interactions between the annular space and tubing which leads
to the unstable behavior (heading limit cycles) at low and
intermediate gas injection rates. In addition it is necessary that
the model becomes stable at high gas injection rates. The idea
is to use a simple model basically relying on three differential
equations conserving mass in the tubing and casing, and a
couple of algebraic equations (of state) for approximating
energy and impulse balances. At the cost of a more
complicated, yet accurate, model, differential equations
describing energy balances and impulse balances may also be
included. To sum up, the nonlinear dynamic gas lifted well
model consists of:
• Model of the pipes (casing and tubing):
1. Three ordinary differential equations conserving
masses in casing and tubing.
2. Algebraic equations (of state) relating pressure,
temperature, and liquid and gas holdup to each other
in casing and tubing.
3. Algebraic equations for pressure head.
• Model of gas injection choke: An algebraic equation
describing the relation between the pressure upstream and
downstream the gas injection choke and the mass flow
rate through the choke.
• Model of the gas injection valve: An algebraic equation
describing the relation between the pressure upstream and
downstream the gas injection valve and the mass flow rate
through the valve. The equation will vary depending on
the type of gas injection valve used.
• Model of the production choke: An algebraic equation
describing the relation between the pressure upstream and
downstream the production choke and the mass flow rate
of gas and liquid through the choke
The advantages with this simple dynamic model structure
are many. Compactness is one appealing feature (only a set of
ordinary differential equations and algebraic equations).
Secondly, it is able to capture the main dynamic behavior of
5
gas lifted well both at low, medium (unstable operating
conditions) and high (stable operating conditions) gas
injection rates. Even so important, the model may easily be
linearized, meaning that it is suitable for linear controller and
estimator design. In addition, the parameters in the model can
be tuned so that the model fits measured real time-series of
pressures, temperatures, and flow rates from a gas lifted well.
Fig. 11 shows an open-loop simulation where the simple
model structure has been used to model the unstable gas lifted
well in Fig. 5. The model has been implemented and simulated
using MATLAB. MATLAB (see e.g. Ref. 5) is an integrated
technical computing environment that combines numeric
computation, advanced graphics and visualization, and a high-
level programming language, and provides a perfect
environment for dynamic analysis and model-based controller
development and design. As seen from the simulation results,
the model is able to capture the oscillating limit cycle
conditions, also known as casing heading, in the experimental
gas lifted well.
Linear Models For Controller Design
A nonlinear dynamic gas lifted well model in accordance with
the proposed structure may be used directly as part of the
model-based stabilizing gas lift controller shown in Fig. 10.
However, it is sometimes difficult to design a model-based
controller based on a nonlinear model. The preferred way for
utilizing the derived nonlinear model will therefore be
linearization. To locally capture the dynamic behavior of an
unstable operating point of a gas lifted oil well, the nonlinear
model in accordance with the structure described above may
be linearized in the current operating point of interest.
Representing the local dynamics of a gas lifted well using a
linear state-space model or, equivalently, a transfer function
model then locally captures the dynamic behavior in the
neighborhood of an unstable operating point. Several authors
have previously used transfer function models for locally
representing the dynamics of gas lift wells in order to derive
stability criteria (see Refs. 4, 7 and 8).
We have developed two efficient ways of generating these
kind of linear gas lifted well models. One way is, as already
alluded to, by (numerical or analytical) linearization of a
nonlinear dynamic gas lifted well model in accordance with
our invented structure described above. Another way of doing
it is by closed-loop identification experiments on a gas lifted
oil well modeled in OLGA where the closed-loop system is
stable in the operating point in question.
Used in combination with advanced techniques from
control theory (see e.g. Ref. 9) the linear local gas lifted well
models (as described above) can be used to design model-
based linear locally stabilizing gas lift controllers. In this way,
an (optimal) operating point that is unstable in open-loop (i.e.,
without active control), becomes locally stable in closed-loop
(i.e., when the stabilizing gas lift controller is actively used).
In order to generate globally model-based stabilizing gas lifted
well controllers, we combine the model-based linear locally
stabilizing gas lifted well controllers described above. Each
model-based gas lifted well controller will then consist of a
6 B. Jansen, M. Dalsmo, L. Nøkleberg; K. Havre, V. Kristiansen, P. Lemetayer
family of model-based linear stabilizing controllers, each of
which will be valid in a predefined neighborhood of an open-
loop unstable operating point, and switching between the
controllers will occur based on predefined logical rules. An
illustration of this concept is shown in Fig. 12.
Possibilities for Different Control Structures
A nice feature of the model-based controller concept is its
flexibility concerning the selection of the particular controller
structure to use, i.e., which measurements to use for control
and estimator purposes and which chokes to use for active
manipulation. The pairing of measurements and manipulation
chokes may vary from well to well, and therefore it is
important to develop a concept for control that is robust
enough to tackle such differences. Several control structures,
in line with our general concept shown in Fig. 12, have been
simulated based on a software link between MATLAB and
OLGA. The gas lifted well is then modeled in the multiphase
simulator OLGA and the model-based gas lift controller is
implemented in MATLAB, and the experiment itself is run in
the MATLAB environment. What is observed in all the
simulations of the various model-based controller concepts is
that the heading phenomena is eliminated through the active
and continuous manipulation of the opening of the production
choke and/or the opening of the gas injection choke.
Surprisingly, even rather simple control structures are able to
cope with the instability. The reason for this is the model-
based feature of the controller concept, making the controller
able to estimate what is not measured. Indeed, using e.g. only
measurements of pressure in the production tubing as input to
a model-based gas lift controller, the gas lifted well may be
stabilized only through dynamic manipulation of the gas
injection choke. Pressure in production tubing may be
measured anywhere between the bottom of the well to the
wellhead. Fig. 13 shows a model-based controller structure
using measurements of wellhead pressure to manipulate the
gas injection choke, and Fig. 14 shows a corresponding
simulation result for the well in Fig. 5 using the
MATLAB/OLGA-link. What is easily seen from Fig. 14 is the
efficient way of removing the heading as soon as the controller
starts.
Similarly, using only measurements of pressure in casing
as input to a model-based gas lift controller, the gas lifted well
may be stabilized only through dynamic manipulation of the
production choke. A controller structure using this
measurement is shown in Fig. 15, and Fig. 16 shows a
corresponding simulation result for the well in Fig. 5 using the
MATLAB/OLGA-link.
Conclusions
Field experience and simulations demonstrate that more oil
can be produced with less lift gas provided that automatic
control is applied for stabilization. The production increase is
particularly prominent on high PI wells or when having high-
pressure lift gas. Stabilisation with enhanced control avoids
using lift gas in an inefficient way or introducing continuously
a high back-pressure.
SPE 56832
Enhanced control of gas lifted wells also gives other
benefits. It allows avoiding gas-lifted wells instability under
low lift gas injection thus making it feasible to reduce the gas
injection rate below the point where instability usually occurs,
and it reduces the effect of disturbances.
The sequenced-based automation is widely field proven to
stabilise production and is extended for other applications such
as ESP and deep offshore risers.
The model based controller described in this article has a
potential for improving the control method developed and
proved by Elf, in that it on-line accounts for well
characteristics and might stabilise more unstable operating
points.
References
1. Avest, D. ter., and Oudeman, P.: "A Dynamic Simulator to
Analyse and Remedy Gas Lift Problems", paper SPE 30639
presented at the 1995 Annual Technical Conference &
Exhibition, Dallas, Oct. 22-25.
2. Bendiksen, K.H, Malnes, D., Moe, R, and Nuland, S.: "The
Dynamic Two-Fluid Model OLGA: Theory and Application",
SPEPE, May 1991, pp. 171-180.
3. Lemetayer, P, and Miret, P.M.: " Tool of the 90's To Optimize
Gas-Lift Efficiency in the Gonelle Field, Gabon", paper SPE
23089 presented at the Offshore Europe Conference, Aberdeen,
Sept. 3-6, 1991.
4. Garnaud, F., Casagrande, M., Fouillout, C., Lemetayer, P.: "New
Field Methods for a Maximum Lift Gas Efficiency Through
Stability", paper SPE 35555 presented at the 1996 European
Prod. Opr. Conf., Stavanger, April 16-17.
5. Redfern, D., and Campell, C.: The MATLAB 5 handbook,
Springer, New-York, 1998.
6. Blick, E.F., Enga, P., and Lin, P.: "Theoretical Stability Analysis
of Flowing Wells and Gas-Lift Wells", SPEPE, Nov. 1988, pp.
508-514.
7. Alhanati, F.J.S., Schmidt, Z., Doty, D.R., and Lagerlef, D.D.:
"Continuous Gas-Lift Instability: Diagnosis, Criteria, and
Solutions", paper SPE 26554 presented at the 1995 Annual
Technical Conference & Exhibition, Houston, Oct. 3-6.
8. Tinoco, M.M.:"Validation and Improvement of Stability Criteria
for Gas-Lift Wells", Artificial Lift Projects Research Report,
Department of Petroleum Engineering, The University of Tulsa.
9. Skogestad, S., and Postlethwaite, I.: Mulivariable Feedback
Control, John Wiley & Sons, 1996.
UNSTABLE STABLE
Production
Region of optimumlift gas utilisation
Gas Injection
Fig. 1—Typical lift performance relationship curve for a gas lifted
well.
SPE 56832 AUTOMATIC CONTROL OF UNSTABLE GAS LIFTED WELLS 7

Wellhead Rate
Wellhead
Pressure Temp.
Meas. O il p r o d u ctio n r ate v e rs u s tim e
(s im u la te d b y O L G A )
Production Choke O il p ro d uc tio n ra te (k g /s ) G as lif t rate = 1 .2 0 kg /s
10 F lo wing b o tto m ho le p re s s ure (b a r) 94

Gas/oil Separator 9 .9 9 3 .8

O il p ro d u c tio n ra te (k
9 .8 9 3 .6

P re s s u re (b a
9 .7 9 3 .4

9 .6 9 3 .2
Lift gas Gas from compr. 9 .5 93
Gas Injection Choke 9 .4 9 2 .8

9 .3 9 2 .6
Casing 9 .2 9 2 .4
Head
Pres 9 .1 9 2 .2

9 92
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0

T IM E (h )
Annulus
Fig. 4—Unstable production for a gas lifted well – constant gas lift
rate = 1.2 kg/s.
Production tubing

Gas Lift Valve 12

Region of optimum operation
Bottomhole
Pressure 10

Oil production rate (kg/s)

8

Reservoir 6
Oil production rate - Stable flow

Fig. 2—A schematic of a gas lifted oil well including the most 4 Oil Production rate - Unstable flow
common measurements
2

0
O il p ro d u c tio n r a te v e rs u s tim e 0.0 0.5 1.0 1.5 2.0 2.5
Gas lift rate (kg/s)
(s im u la te d b y O L G A )
O il p ro d uc tio n ra te (k g /s ) G as lift rate = 0 .6 0 kg /s Fig. 5—Lift performance relationship for a typical gas lifted well.
28 F lo wing b o tto m ho le p re ss ure (b a r) 150 The solid curve corresponds to injecting lift gas directly through
24 140 the lowest gas injection valve at constant rate, and the dotted
P re s s u re (b a r)
O il p ro d u c tio n ra te (k

curve shows the resulting average production when the lift gas is
20 130
injected at constant rate through the gas injection choke.
16 120

12 110

8 100 OIL CHOKE

4 90

0 80
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 OIL
T IM E (h ) SYSTEM
MW
MONOWELL U E
Fig. 3—Unstable production for a gas lifted well – constant gas lift SEQUENCES L L
rate = 0.6 kg/s. T L
I S
LIFT GAS
SYSTEM

LIFT GAS
CONTROL VALVE

Fig. 6—System overview.

8 B. Jansen, M. Dalsmo, L. Nøkleberg; K. Havre, V. Kristiansen, P. Lemetayer SPE 56832

100
OIL INCREASE External
Reference
GAS DECREASE

STABILITY
%
Production
Model-based Choke Gas Lifted Process
Controller Oil Well Measurements
Standard well Gas Injection
Gas-lift optimisation & automation Choke
operation
0

4 5 6 7 8 9 10 11 12 13 14 15 16
Time(hours) Model-
based
Fig. 7— Example of how the automation concepts influence well Estimator
behavior.
Fig. 10—A schematic of the new model-based stabilizing gas lift
controller structure.
.
Operator stations

40
Oil production rate

computer network 30

local
controller
kg/s
20

10

0
0 1 2 3 4 5 6 7 8

time[hour]
Fig. 11—An open loop (no active control) transient simulation for
an experimental gas lifted well based on a simple nonlinear gas
lifted well model.

Reference
Controller I Production
Choke Gas Lifted Process
Controller II Oil Well Measurements
Gas Injection
Choke
Controller n

Fig. 8— System architecture

Estimator

Fig. 12—A model-based stabilizing gas lift controller consisting of

THP THT
several linear locally stabilizing gas lifted well controllers.

MONOWELL LOGIC OIL PRODUCTION CHOKE

& PARAMETERS IGR CHP
Wellhead Pressure Opening of Measurement of
Reference Model-Based Gas Injection Choke Wellhead Pressure
SSV OIL NETWORK Gas Lifted
Gas Lift
Well
Controller
GAS NETWORK

FCV SDV
GAS-LIFT INJECTION VALVE

Fig. 13—Control structure for stabilization of a gas lifted well

using measurements of wellhead pressure and a model-based
Fig. 9: Instrumentation overview controller for manipulation of the gas injection choke. The
numbers in brackets refer to Fig. 2.
SPE 56832 AUTOMATIC CONTROL OF UNSTABLE GAS LIFTED WELLS 9

Wellhead pressure
Total oil production rate
25
50

40

30
Controller Startup
bar

kg/s
20 20
Controller Startup

10

15 -10
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
time [hours] time [hours]

Fig. 14—Stabilization of wellhead pressure (left) and stabilization

of the oil production rate (right) using the control structure in Fig.
13.

Measurement of
Opening of Casing Pressure or
Reference Pressure
Model-Based Production Choke Pressure in Gas Supply Line
Gas Lifted
Gas Lift
Well
Controller

Fig. 15—Control structure for stabilization of a gas lifted well

using measurements of casing pressure and a model-based
controller for manipulation of the production choke. The numbers
in brackets refer to Fig. 2.

Casing Pressure
Total oil production rate
140 60
Controller Start
120 40
kg/s

20
bar

100

80 0

60 -20
0 5 10 15 0 5 10 15
time [hours] time [hours]

Fig. 16—Stabilization of casing pressure (left) and stabilization of

the oil production rate (right) using the control structure in Fig.
15.