SPE 125050

Comparing the Benefits: Use of Various Well Head Gas Coning Control Strategies to
Optimize Production of a Thin Oil Rim
E.D. Nennie, SPE, S.V. Savenko, G.J.N. Alberts, SPE, M.F. Cargnelutti, TNO the Netherlands, and E. van Donkelaar,
Shell International Exploration and Production.

Copyright 2009, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE 2009 Annual Technical Conference and
Exhibition held in New Orleans, 47 October 2009.
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, Texas 75083-3836 U.S.A., fax 01-972-952-9435.

Abstract
With an increasing number of smart well applications being
installed in the field, more knowledge is required to optimize
their operation. This paper compares the benefits of various
wellhead gas coning control strategies to optimize production
of a thin oil rim. This study is performed within the
"Integrated System Approach Petroleum Production (ISAPP)"
knowledge center of TNO, TU Delft and Shell.
For this study a field case model is used, which has been
validated with field data. The field case is a thin oil rim with a
horizontal well. Due to the location of the horizontal well in
the oil rim, the well is particularly susceptible to gas coning.
Besides gas coning, wax precipitation is a second production
constraint. This makes this well challenging to operate.
Different production strategies are investigated and compared
against each other: intermittent production and continuous
production with pressure differential control.
The results of the different production strategies are
presented by analyzing the advantages and disadvantages for
the different gas coning control strategies, satisfying the given
constraint of gas influx. This study reveals the difference in
the cumulative production between the two strategies. The use
of a closed loop control strategy can lead to a larger oil
production in the same amount of time.
This paper shows the viability of using dynamic simulation
models to quantitatively assess the benefits of various
production optimization strategies. This allows operators to

compare emerging smart well technologies, and increase trust

in specific technologies that could be of an added value to
their operation. Even though much has been published about
the potential benefits of a smart field philosophy, few
published field cases are available. This paper offers a field
case testimony of the comparison of various feedback control
strategies for purpose of production optimization.
Introduction
With increasing knowledge and improving technologies, more
complex reservoirs (with respect to location and dimensions)
can be explored and produced. This brings new challenges in
exploration, drilling and production. Furthermore, existing
reservoirs require new insights to be able to increase ultimate
recovery. Dedicated simulation software tools can offer these
new insights by helping to understand production instabilities
and test new control strategies to avoid instabilities and to
optimize production.
The field under investigation has most of its wells drilled
with long laterals in a thin oil rim, making them particularly
susceptible to gas coning. Gas coning is a phenomenon where
the gas oil contact of a reservoir slowly moves towards a well
as a result of high drawdown. Eventually, the free gas is being
drawn into the well, see Figure 1. Furthermore, the reservoir
temperature is low enough to cause wax deposition.
At high production rates, a well will suffer a large gas
influx, which cannot be handled by the topside equipment. For
low production rates, a well will suffer increased wax
deposition due to the lower fluid temperature [Nennie, 2008].
Therefore, due to gas coning and wax deposition, some of the
wells are operated intermittently. Goal of this study is to
determine whether instead of the intermittent production
continuous production is more beneficial and if so, quantifying
the difference between these control strategies.
Leemhuis et al. already showed that gas coning can be
controlled by changing the wellhead choke [Leemhuis, 2007].
A PID controller was used to bean back the choke in order to
keep the gas fraction at a constant level. Leemhuis et al. also

SPE 125050

found for their test case an optimal gas fraction set point
which yields a maximum in oil production. For this paper a
similar control strategy was implemented in a field case,
taking into account practical issues such as uncertainties in the
gas fraction measurement at the wellhead choke.

Figure 2

Figure 1 Snapshot of free reservoir gas drawn into

target well in a thin oil rim
The evaluation of the two production strategies is done in a
fully coupled well-reservoir simulation. The coupled dynamic
well-reservoir simulator, developed within the research
framework of ISAPP, is used to couple a dynamic reservoir
model to a dynamic well model of one of the horizontal wells
located in the field. The fully coupled system is required as for
certain dynamical phenomena the well dynamics are
influenced by the reservoir dynamics and vice versa [Nennie,
2007]. The pressure transients during gas breakthrough are
much faster and closer to reality for the coupled simulation
than calculated with a stand-alone reservoir model.
Model Description
Dynamic Reservoir Model
To simulate the reservoir, MoReS is used. This is a reservoir
simulator capable of handling fractured and non-fractured
reservoirs. It was designed for a wide range of simulation
applications.
The reservoir model is a sector model of the complete
field, given in Figure 2. The figure shows the target well;
several wells in the vicinity have been omitted for clarity.
These wells are included in the numerical model. The field
case is a thin oil rim containing three segregated phases:
water, oil and gas. The thin oil layer contains a nearly
horizontal well with the perforated part of about 300 meters.

Target well and the producing field

The model of the field has a horizontal extent of 2618m by

5876m, with a thickness at the location of the well being equal
to 96m (from 1017m to 1113m below the surface). The initial
gas oil contact level is at 1063m, and oil-water contact is at
1110m below the surface. Average porosity of the rock equals
0.25 and isotropic permeability 215 mD. Reference pressure is
1.21x107 Pa and volumes initially in place are correspondingly
water: 8.46x104 MMSm3, oil: 14.40 MMSm3, and gas:
3.58x107 MMSm3. Two wells are situated in a close vicinity of
the target well. Production of those wells has a significant
influence on performance and thus the model should
reasonably reproduce it as well.
The reservoir has a fixed volume with no flow boundary
conditions at the far field boundary and bottom hole pressure
constraints as a boundary condition for the well. The model
contains 10904 active grid blocks (791 aquifer blocks), with
typical block dimensions of approximate 100x100x10m. In the
near-wellbore region around the well, three levels of
refinement were applied to allow for a faster response. The
reservoir is fully connected, i.e. no faults are present, while the
interaction with the adjacent gas reservoir is taken into
account by adjusting the effective grid block volume values at
the boundary. Hysteresis in the oil-gas relative permeability
table is included for the volume below initial gas oil contact
with the critical and residual gas saturations correspondingly
0.10 and 0.30. Critical water saturation is 0.23 for the
complete model while the Corey exponents for gas and water
phases are set equal to 3.00, and gas oil and water oil
exponents to 1.50. Residual water saturations of gas and water
are 0.20 and 0.30 correspondingly.
The reservoir model has been history-matched with
production data of the complete sector since it came into
operation. The historical production data is given as average
production over 30 days. Since almost no water is produced
during the validated period, the most sensitive quantity to
monitor is the gas-to-oil ratio (GOR), see Figure 3.

SPE 125050

the well model, resulting in a discrete inflow pattern. The

number of reservoir grid cells intersecting the well determines
the number of inflow sources. The grid cell size in the
horizontal section of the well model is 3m. In the vertical
section the grid cell size increases up to 25m at the wellhead.
In the simulation model the flowline pressure is taken as a
constant boundary condition of 1.4x106 Pa. The mass inflow
of gas, oil and water is the boundary condition at the bottom
hole of the well. Between the wellhead and the flowline a
choke is installed with a maximum diameter of 0.03m. The
reservoir temperature is assumed to be constant at 44 C. The
well model takes into account heating and cooling of the
mixture along the wellbore due to heat transfer, evaporation,
etc.
Figure 3

Simulated GOR dynamics vs. the history data

In Figure 3, the simulated data (red line) and historical data

(green line) of the monthly averaged GOR is plotted, together
with the instantaneous simulated GOR (blue dots). In the
reservoir model, relative permeability, aquifer strength, and
KDH factor are set such to provide the best match to the
historical data. The same set of parameters provides a very
good match for the production of adjacent wells.
Dynamic Well Model
For the well, the commercial tool OLGA is used. OLGA is a
one-dimensional multi-phase flow simulator that can be used
to determine the impact of well/pipe design on the flow of oil,
water and gas. It is capable of simulating transient flow
behavior of the well from bottom hole to the wellhead for all
flow regimes.

Figure 5 Wellhead pressure versus the time of the

measurement for different values of Cd {(a) 0.92; (b) 0.60}

Figure 4

Sketch of the target well

The target well has a true vertical depth (TVD) of 1088m

below surface and has an along hole depth (AHD) of 1734m.
A well sketch is given in Figure 4. The inner casing is run
until 1457m AHD. From there the horizontal section is open
hole. The tubing has an inner diameter of 0.076m and is a
slotted liner in the horizontal section.
The horizontal section of the well intersects the reservoir
model in the oil rim. The reservoir model has a refined grid in
the near-wellbore region. For each reservoir grid block the
well has an inflow source assigned to the nearest grid cell in

From the beginning of 2007 until the end of 2008 fourteen

pressure and flow measurements were taken from the target
well. For a certain measurement point of the well test data, the
discharge coefficient Cd of the wellhead choke is matched
such that the wellhead pressure corresponds to the test data.
With this value of the discharge coefficient the other data
points are computed. In Figure 5(a) the wellhead pressure is
given for Cd=0.92. For this discharge coefficient the wellhead
pressure of the latest data point is matched. For the data points
beyond the first of January 2008 the wellhead pressures are in
very good agreement with the well test data, but before this
date the error between well test data and simulated wellhead
pressure is significant. Even though the absolute value is not
correct the trend is the same as observed from well test data.

To understand the absolute error in wellhead pressure a

second value of Cd is evaluated where the discharge
coefficient is matched to the first data point, see Figure 5(b)
where Cd=0.6. The opposite behavior is observed as in Figure
5(a). For both values of Cd the simulated trend in wellhead
pressure is similar as the well test data. Possible explanations
for the mismatch in wellhead pressure can be that the wellhead
choke has been changed or replaced in the beginning of 2008,
or some uncertainties in the model e.g. wax deposition.
Coupled Dynamic Model
The coupling between OLGA and MoReS has been done in
Matlab. OLGA has a Matlab toolbox which makes it possible
to exchange data with OLGA during the simulations. MoReS
communicates with Matlab via file I/O. The interface between
the well and the reservoir is the natural boundary between the
two models. Each inflow point is modeled as a separate
producer in the reservoir simulator. At the predefined 32
producers the well simulator feeds the reservoir simulator with
the computed bottom hole pressure per producer (inflow point)
at a given exchange time step. At the same moment in time the
reservoir simulator feeds the well simulator with the mass
flow, gas fraction and water fraction per producer. Because the
coupling of the models is explicit, stability issues limit the
exchange time step and therefore CPU time can be a
challenge. To limit stability problems, a staggered grid is
chosen, based on an analogy with fluid-structure interaction
simulations, see Figure 6 [Rixen, 2003]. The staggered grid is
first-order accurate in time.

SPE 125050

boundary condition replacing the well model, until the wellhead choke setting changes again etc.
Control
Feedback control is essentially a way to bring a measured
variable in a process to a certain reference value by means of a
controller device. In reservoir management literature feedback
control is sometimes referred to as active control as
opposed to pro-active control. In this paper, feedback control
is used to investigate the possibility of keeping a well at a
constant gas rate and in this way preventing full gas
breakthrough.
For this investigation, a feedback controller is
implemented on the coupled dynamic model. This controller is
designed to modify the choke opening at the wellhead to keep
the gas volume flow at a constant rate by measuring the
pressure drop across the wellhead choke.

Figure 7

Feedback controller scheme

Figure 7 shows the setup in a schematic way. The time

varying gas volume flow y (t ) in the wellhead is used as a
feedback signal. Upon comparing y (t ) to its set point r the
error signals e(t ) is obtained:

e(t ) = y (t ) r

(1)

The error signal, which is a direct measure of how far the

gas volume flow is away from its set point, is used as the input
of controller K (e(t ) ) . The output u (t ) of the controller is
continuously updating the wellhead choke setting in such a
way that the gas volume flow will converge to its set point:

u (t ) = K (e(t ) )
Figure 6 Staggered grid used in coupling well and
reservoir model [Rixen, 2003]
The coupled simulator has the same boundaries as the
individual well and reservoir simulators, i.e. wellhead choke
opening and no-flow boundary conditions at the far field. The
main benefit of the coupled model is that the interface
boundary conditions at the horizontal well part are replaced by
a dynamic well boundary for the dynamic reservoir model and
vice versa.
The controlled simulations are performed in a semicoupled way: when the wellhead choke has changed, the
simulations are performed with the coupled simulator, as
described above until a semi-steady-state is reached. This is
the timescale where well-reservoir interaction is of importance
[Nennie, 2007]. After the coupled simulator has reached the
semi-steady-state, the simulation is continued with only the
reservoir simulator with a constant bottom hole pressure as a

(2)

In this investigation, a Proportional Integral (PI) controller

is used for the controller device. A motivation to do so is
because the simplicity and good performance of a PI controller
have made it the standard in many industries. An additional
advantage of a PI controller is the fact that, unlike more
advanced model based approaches, its design asks for little
knowledge about the many variables that drive a gas
breakthrough. For a PI controller the relation between the
controller input and output, which is stated by equation 2, is
given by:
t

u (t ) = K p e(t ) + K I e( )d

(3)

The above formula shows that the output of a PI controller

in this application consists of two terms. The first term is for
proportional feedback of the error signal e(t ) that is scaled by
means of the proportional gain K p . The second term is

SPE 125050

feedback of the sum of all previously measured errors, scaled

by the integral gain K I . The integral term accelerates reaching
of the set point and provides off-set free control.
Given the above, designing a PI controller that suits the
purpose for this investigation comes down to finding the
values for the controller gains K p and K I that make the gas
volume flow converge to the set point and stabilize. This
choice is driven by two opposite yet equally important
boundary conditions: stability and performance. High gains
will generally give fast tracking performance but may yield
instable closed loop systems, whereas controllers that have a
low gain will be stable but show a slow tracking performance.
A method that allows one to select controller gains that suffice
both criteria is given by Ziegler-Nichols [Franklin, 1994]. In
many cases this method is not practical however and, like in
this case, the controller needs to be designed by trial and error.

Figure 8 Oil mass flow versus time for intermittent (red)

and controlled production (green, blue)

Results
Three production strategies are simulated: intermittent
production, controlled production from the beginning (control
1) and controlled production from the moment the intermittent
production reaches its threshold of gas inflow (control 2). Both
control 1 and control 2 are compared with intermittent
production.

Intermittent production
In the field, the target well produces intermittently. The well
produces oil and gas until a given threshold of gas inflow is
reached (determined by the constraints of downstream process
facilities). After the threshold is reached the wellhead choke is
closed. After three weeks of shut in the choke is re-opened.
During the simulated time span no water is produced. The
water is drawn towards the well, but water coning will occur
after several years (not included in this analysis).

Figure 9 Gas mass flow versus time for intermittent

(red) and controlled production (green, blue)

Control 1
In Figure 8 the red line indicates the oil mass flow for one
intermittent cycle. The wellhead choke is closed when the gas
inflow at the well is larger than 1 kg/s. In this case, this
happens after 50 days of production. The green line shows the
controlled production. The controller modifies the choke
setting such that the gas inflow is kept at 1 kg/s. Since at start
up of the well the gas inflow is smaller than the set point of the
controller the choke is opened considerably, resulting in an
initial larger oil production.
Figure 9 shows the gas mass flow versus time. The gas
mass flow of the controlled production is brought to its set
point and more gas is produced compared to the intermittent
production. In this case the controller increases the total mass
flow, but this results in an increase of gas instead of oil flow.
Looking at the cumulative oil produced, in Figure 10, the
difference between the intermittent and controlled oil
production is negligible. Figure 11 shows the cumulative
produced gas for intermittent and controlled production. As
mentioned befor the gas production increase significantly.

Figure 10 Cumulative produced oil [kg] versus time for

intermittent (red) and controlled production (green, blue)

SPE 125050

production. Both strategies produce the same amount of oil for

the same reservoir depletion. The benefit of the controlled
production strategy is that the oil is produced faster. In this
case the same amount of oil is produced in 1/3 of the time.

Figure 11 Cumulative produced gas [kg] versus time for

intermittent (red) and controlled production (green, blue)
Due to larger choke opening, the controlled production is
depleting the reservoir much faster than the intermittent
production strategy. Since the difference in cumulative oil
production is negligible, this simulation reveals that controlled
production is not always a better option.

Figure 12 Cumulative produced oil versus cumulative

produced gas for intermittent production (red), controlled
production (green, blue)
Conclusions

Control 2
The use of the controller from the beginning of the simulations
did not result in larger cumulative oil production, a second
controlled production strategy is tested. The controller is
turned on after 50 days of production, when the intermittent
production strategy closes the well choke. With the control
strategy the gas inflow is restricted to 1 kg/s. From Figure 8 til
Figure 11 the blue line indicates the control 2 results.
The advantage of the controlled production strategy is that
there is a continuous production whereas the intermittent
production strategy has no production during shut in. Figure 9
shows that the controller is able to restrict the gas inflow
around 1 kg/s. The reason of the fluctuation of the gas flow
around the set point is the presence of variable parameters in
the controller, which are assumed constant in the present
model. Even though these parameters need to be adjusted
during the simulations, the results show a possibility of
continuous production.
In Figure 10 is shown that oil is produced at higher rate
with the controlled strategy than with the intermittent
production strategy. Not only oil is produced faster, but also
gas, as shown in Figure 11.
An indication for optimal production in the long term is the
cumulative oil versus cumulative gas production (Figure 12).
As gas is the major driving force for oil production it
determines the rate of depletion. Large production of gas could
lead to fast depletion of the reservoir resulting in smaller total
cumulative oil on the long term.
The black circle in Figure 12 indicates that there is only a
small difference between intermittent and controlled

A coupled dynamic well-reservoir simulator is used to

investigate several gas coning control strategies. This paper
shows that the use of a controlled strategy can result in higher
oil production in a shorter time span. In the present field case
one of the control strategies (control 2) was able to produce
the same amount of oil in 1/3 of the time needed by the
intermittent strategy.
However, it is important to determine the moment the
controller should be used. For example, the first control
strategy shows that the use of a control strategy can lead to a
faster depletion of the reservoir in comparison with
intermittent production.
This paper shows that smart well technologies can actually
lead to an improved oil production. The knowledge gained by
testing different control strategies within the coupled
simulation environment can be used to optimize the ultimate
recovery of thin oil rims.

Future work
In the present analysis just one intermittent production cycle is
presented, representing 85 days of production. To reach a
definitive conclusion about the influence of controlled
continuous production on the long term, more cycles are
required.
Another point of interest is the influence of the variable
parameter in the controller. More study is required to let the
controller work automatically, as it was designed without
human intervention.

SPE 125050

Acknowledgment
The authors would like to thank the Shell operating unit for
giving feedback and all the data required for this analysis;
Walter Renes (TNO Science and Industry) and Gerard Joosten
(Shell International Exploration and Production) for their
feedback and sharing their expertise; and the ISAPP
knowledge center for making this project possible.

References
Franklin, G.F., Powell, J.D., Emami-Naeini, A., Feedback Control
of Dynamic Systems, Addisom-Wesley Publishing Company,
New York, 1994
Leemhuis, A.P., S.P.C. Belfroid, G.J.N. Alberts, Gas Coning
Control for Smart Wells, SPE paper 110317 presented at the
2007 SPE Annual Technical Conference and Exhibition,
Anaheim, California, U.S.A., 11-14 November
Nennie, E.D., Alberts, G.J.N., Belfroid, S.P.C., Peters, E., Joosten,
G.J.P., An Investigation Into the Need of a Dynamic Coupled
Well-Reservoir Simulator, SPE paper 110316 presented at the
2007 SPE Annual Technical Conference and Exhibition,
Anaheim, California, U.S.A., 1114 November 2007
Nennie, E.D., Alberts, G.J.N., Peters, E., van Donkelaar, E., Using a
Dynamic Coupled Well-Reservoir Simulator to Optimize
Production of a Horizontal Well in a Thin Oil Rim., SPE paper
118173 presented at the 2008 SPE Abu Dhabi International
Petroleum Exibition and Conference, Abu Dhabi, U.A.E., 3-6
November 2008
Rixen, D.J., Fluid-structure interaction: An introduction to numerical
coupled simulation, Lecture Notes Faculty of Design,
Engineering and Production Engineering Mechanics
Dynamics, 2003