Gas Production Gathering Network

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

"Gas Production Gathering Networks - A Coupled Surface/Subsurface Approach.

"
Abstract
A typical approach taken to modeling gas field production gathering networks is to input "fixed" flow rates at a
particular point in time to size pipelines and/or identify restrictions to flow. A numerical reservoir simulation
model is often used to forecast the production rates for each well in the field. The flowing conditions in the surface
gathering network can vary significantly over the life of reservoir due to the field development strategy being
pursued, pressure depletion of the reservoir, and the volumes of the gas and liquid that are produced. Until
recently, the best means available to account for changes in flow rate potential or pressure losses in the reservoir
and surface gathering network has either been to use a "simplistic" description of the reservoir inflow performance
coupled to a "simplistic" surface network or, to run the reservoir model and pipeline model separately, manually
exchanging pressures and flow rates data in a "trial-and-error" fashion until the models converge.
This paper summarizes the recent "movement" by reservoir simulation software vendors to allow their models to
couple to 3rd party pipeline and reservoir simulation models. The Sexsmith Gas Condensate Field located in
Alberta, Canada is used to illustrate how these integrated" models can more closely approximate the production
characteristics of the reservoir and act as a focal point for planning and management of the surface gathering
network and the transportation pipelines located downstream. The paper also offers an opinion of where further
effort should be focused to make these modeling tools more useful to the end user.

Introduction
The design of a multiphase flow gathering network requires an estimate of flow potential from each well. Rarely is
the aerial extent of the reservoir so well defined at the outset that all potential well locations can be identified. The
flow potential of the fluid phases produced up the wellbore into the gathering network vary between wells
depending upon the characteristics of the reservoir, and pressure depletion or drive mechanism of the reservoir.
Accordingly, the efficient transport of reservoir fluids through the gathering network is usually difficult to predict
over the producing life of the wells.
The forecast of production rates from wells using a numerical reservoir simulation model is likewise dependent
on backpressures caused by pressure losses in the wellbore and surface gathering network. The historical approach
taken by the reservoir engineer has been to prepare flowing bottomhole pressure versus flow rate (VFP) look-up
tables to approximate the backpressure caused by the wellbore and surface network. Each table is unique with
respect to tubular dimensions, lengths, etc. When dealing with multiphase flow many permutations (e.g. different
water-gas ratios, condensate-gas ratios) are often required to construct a VFP table over potential range of flowing
conditions. Such an approach is suitable for single well, single flow line gathering networks however, such
configurations are not typical of most pipeline networks. It is not uncommon for networks to include numerous
wells, tubing descriptions, pipeline branches and loops, as well as a variety of surface equipment such as pumps,
compressors, line heaters, separators, etc.
Attempts by the reservoir engineer to forecast the flow rate potential of the reservoir, and the pipeline design
engineer to design a network to efficiently transport reservoir fluids is particularly difficult in high rate, low
pressure systems. Simplification of the hydraulic component of the gathering network or the reservoir model
component usually introduces an unacceptable error to engineering calculations. The use of conventional surface
network models, that use one-dimensional, tank-type reservoirs, may address the backpressure problem, but they
cannot accurately forecast transient production profiles. This is especially true in low permeability, hydraulically
or naturally fractured reservoirs, or those reservoirs exhibiting multiphase flow effects.
The problems described have long been understood as an impediment to improving the accuracy of reservoir
simulation forecasts and the design of pipeline networks to transport the produced fluids. The solution to the
problem demands integration of various engineering disciplines, and their software technologies.
Recently, some vendors of reservoir simulation software have created interfaces allowing third party pipeline

network models to connect with their software. The coupling of the models has allowed the aforementioned
problems to be more rigorously modeled. However, in the process of coupling and running the models, it has
become very apparent that in pursuing integrated solutions that both engineers and the software developers
maintain a close dialogue.
Field Description
The sour gas reserves in the Montney formation found near Sexsmith (see Figure 1), in northwestern Alberta,
Canada represents one of the largest gas discoveries in the past 20 years. The first wells were drilled in the field in
1992. As additional wells were drilled, completed and tested, it soon became apparent that there was a need for a
major gas processing facility. As Conwest Exploration Company Limited held the largest gas reserves in the area,
they undertook to design and construct the new facilities. (In 1996, Conwest merged with Alberta Energy
Company Ltd. and now forms part of the AEC Oil and Gas Partnership). Key to the design of the facilities was the
evaluation of the pool reserves and the preparation of a gas deliverability forecast. As interest in the area
heightened, still more wells were drilled and additional gas reserves were found in the Montney formation in the
nearby Valhalla and Knopcik areas. Incorporation of these reserves resulted in the construction of a somewhat
larger plant than initially anticipated, with a nominal capacity of 200 MMscf/d (5635 E+03 m3/d).
The gas pipeline network connecting the various gas wells to the Sexsmith gas processing facility extends eastwest over six townships (36 miles/58 kms), and north-south over four townships (24 miles/38 kms), as shown in
Figure 2. The western portion of the network is also connected to the AEC Hythe plant, which can be used to
process a portion of the gas production.

Sexsmith Field Modeling History


After discovery and the initial delineation drilling were complete, a description of the reservoir evolved that
allowed construction of a numerical reservoir simulation model. The short-term objective of the model study was
to determine the deliverability potential of the wells drilled in the Sexsmith Montney sour gas pool. To generate
an estimate of flow rate for each well, the reservoir simulator was run employing flowing bottomhole pressure
versus flow rate look-up tables with a common tubinghead pressure. During each timestep taken by the reservoir
simulation model, the flow rate was used to interpolate a flowing bottomhole pressure from the VFP table at the
assigned wellhead pressure. The adverse consequence of such a simplifying assumption was immediately obvious,
since any change to the well completion, produced fluid properties, choke setting or surface pipeline network
modification could alter the wells flow rate potential. It became obvious that the information with which to plan
the long term drilling development of the field, and the pipeline gathering network necessary to sustain a
production rate plateau carried an large component of uncertainty.
Construction of sour gas plants in Alberta is highly regulated and the Sexsmith plant was conceived as an area
facility, open to many participants. As such, the timing of additional wells or facilities may or may not be readily
apparent to the plant operator. An area model is extremely useful in determining changes required to
accommodate additional production volumes due to new well tie-ins or the addition of field compression. While
there was sufficient confidence in the gas deliverability potential over the short term to proceed with construction
of a gas plant, there was less confidence in how the flow rate potential could be sustained over time. Critical to the
long term potential was the timetable for drilling and connecting additional wells to match the plant process
capacity.
The Montney reservoir from which the wells produce has limited flow potential and must be hydraulically
fractured in order for wells to reach their full flow rate potential. When the wells are initially placed on
production, their flow rate potential soon declines to some lower stable rate of production. The rate at which
pressure declines and flow rate changes from each well is a function of gas demand and the unique characteristics
associated with the wells. Over-estimation of the flow rate potential from the wells during the year can result in a
shortage of gas at the process facility, since up to 3 months may be required to drill, complete and connect a well.
Drilling and connecting more wells than necessary will result in a surplus of gas, indicating an unnecessary preinvestment of capital. Both situations are undesirable.
To answer the very real problems of facility planning, completion design, and reservoir management, a more
sophisticated tool was required to model the gas production delivered to Sexsmith Plant. One approach considered
was to run the pipeline and reservoir simulation models independently over a limited prediction period. In such a
scenario, the tubinghead pressure would be pre-assigned to each well in the reservoir simulation model and the
model used to calculate the wells flow rate potential. The flow rate would then be used as an input to the pipeline

network simulation model, and a tubinghead pressure calculated at each well. Flow rates and pressures would be
exchanged manually between the two models in a trial-and-error fashion until the calculated pressures and flow
rates converge to within an acceptable tolerance of accuracy. Previous experience applying this methodology to
other reservoirs, has revealed that the timestep interval must be kept small if the reservoir is depleting rapidly (i.e.,
gas storage reservoirs), or, if physical changes are planned for the wells or surface gathering network. Strict
conformance to the procedure described usually requires such a large number manual exchanges of data to be made
that inevitably timestep lengths are increased to minimize the effort. The pursuit of such approximations can often
cause irregularities to the production forecast and result in misleading information for reservoir management and
the design and scheduling capital and operating investment for pipelines and surface facilities.
Due to the variety of possible development scenarios to be studied at Sexsmith, it was decided that an integrated
reservoir, and wellbore/pipeline network simulation model would provide a much improved solution. Discussion
of the benefits of integrated models date back several years with one of the earliest papers written by Startzman et
ali, published in 1977. Historically, the development of such models has been hampered by the demands that these
models make on the computer processing unit (CPU) and memory requirements. Due to these limitations, the
early models tended to simplify the data input description and limit the modeling options available to the user. A
number of papersii,iii,iv have been written discussing integrated models but these models were often proprietary inhouse models designed to model the specific circumstance of a particular field.
Gas produced from wells connected to the Sexsmith gathering network includes some hydrocarbon liquid and
water that prevents using a dry gas model. Wells are often choked and have line heaters and/or dehydration
equipment installed which makes pipeline heat loss calculations important to fluid phase and pressure loss
calculations. Compressors are also located at various points in the network to reduce the backpressure at the
wellhead to maximize production. Together, these and other limitations listed below eliminated the commercial
integrated modeling packages that AEC was aware:

the fluid property description allowed only a simplified fluid property data input and could not model
multiphase flow;
the reservoir simulation model possessed only a limited number of modeling options in terms of well
grouping and contracts, or were extremely cumbersome to evaluate;
the modeling of the wellbore and surface gathering network had to be performed using a limited list of
vertical and horizontal flow correlations;
the linked models possessed no capability to implement well workover, recompletion, or infill drilling
strategies without stopping the models to redefine changes to the physical system;
the wellbore/surface network models possessed limited flexibility to define equipment installed in the
wellbore;
the surface network simulation model was capable of modeling only simplified gathering network
configurations; and,
the pipeline network simulation model included only a limited list of surface facility equipment types with
limited user defined description characteristics.

AEC was encouraged that several commercially available reservoir and pipeline network packages did not have the
same limitations noted in the integrated models. At some point earlier AEC had had begun dialogue with various
commercial reservoir simulation software vendors and expressed a desire that vendors consider allowing third
party pipeline simulation packages to link to their software. In the intervening period, the vendor of the ECLIPSE
100v black oil simulator had written an interface allowing their model to couple with third party software. A short
time thereafter, Hepgulervi et al described the coupling of this particular black oil reservoir simulator with a
commercial multiphase surface model. Encouraged by this software innovation, AEC asked the vendor of the
FORGASvii gas deliverability and forecasting model to integrate their model with ECLIPSE. A description of
issues associated with this coupling is documented in a paper by Trickviii.
AEC was anxious that the coupling of the two models accomplish the following:
retain the existing functionality found in both simulation models;
allow either model to be run independent of the other and, if desired, linked later with few if any changes
to the data input structure of either model;
be tightly coupled, with data exchanged between the models at the Newton iteration level rather than the
start or end of the timestep to account for transient pressure and flow rate changes within the low
permeability reservoirs at Sexsmith;

seek to minimize the execution time of the integrated model to the greatest extent possible; and,
preserve model stability during calculations.

Model Configuration
The reservoir is modeled using a three-dimensional black oil reservoir simulator. The Open Eclipse interfaceix
uses the Parallel Virtual Machine (PVM) interface to allow the reservoir simulation model to communicate
interactively with the pipeline simulation model. FORGAS is a gas deliverability and forecasting program, used to
model gas pipeline networks including gas-water and/or gas-condensate. When coupled with the reservoir
simulator, the functionality of the pipeline network model can be limited to the calculation of pressure losses or
gains due to pipeline and surface equipment from the reservoir sandface to the delivery or sale point(s), with the
reservoir simulation model controlling target rates. Conversely, the pipeline network model may control targets, in
which case the reservoir simulator provides the reservoir calculation engine. It is also possible to model other
types of reservoirs, for example 1-D tank models, at the same time. This is useful for systems in which other gas
sources exist, but have insufficient data to build a reservoir simulator data set for detailed modeling.
The pipeline simulator uses the PVM interface to start the reservoir program. The PVM interface acts as a
messenger passing information between the pipeline simulator and the reservoir simulator, both of which are
running simultaneously.
The reservoir simulation model computes flow rates. The pipeline simulator uses these flow rates (gas and
liquid) as the basis for calculating pressure changes throughout the gas pipeline network, including the flowing
bottomhole pressure in the wells. These flowing bottomhole pressures are passed to the reservoir simulator
where they are used to compute revised flow rates during the next Newton iteration. The revised flow rates are
then passed back to the pipeline model to re-determine the flowing bottomhole pressure for all wells in the
network. Each time the pipeline model receives a revised flow rate, it compares the value with the previous value.
When the difference in well flow rate between one iteration and the next is less than the flow rate tolerance
specified, the iteration is considered to have converged and the forecast proceeds to the next timestep. If the flow
rate difference for successive calculated values exceeds the specified tolerance, the new flow rate is used to recalculate a revised flowing bottomhole pressure, which is then sent to the reservoir simulator. If convergence is
not reached within a defined maximum number of iterations, the pipeline simulator prints a warning message
indicating the maximum discrepancy, and then proceeds to the next timestep.
If the surface model cannot transport the volume due to some restriction imposed by the facilities, or is
restricted due to contract, the surface model will print a warning message.
After two iterations, the surface model checks to determine whether the reservoir simulator is choking the wells
at the sandface due to either well or group target constraints. The pipeline simulator compares the flowing
bottomhole pressure passed to the reservoir simulator with the pressure value actually used by the reservoir
model. When the difference between these values exceeds a specified tolerance (10 psi or 70 kPa), the surface
model will re-compute the flowing wellhead pressure that corresponds to the restricted flowing bottomhole
pressure. The pressure drop difference, minus a tolerance, is then reported as the wellhead choke.

Integrated Simulation of the Sexsmith Field


As stated previously, the Montney formation is a low permeability gas reservoir and all wells are hydraulically
fractured to enable commercial production. This complicates the simulation of the reservoir due to a long transient
production period for each well. In general, the wells are capable of producing at rates varying from 1 to 6
MMscf/d (28 to 170 E+03 m3/d). Flow potentials after one year of production are typically about 70 percent of the
initial rate.
Through adjustment of wellbore radius to match rate performance a reservoir simulation approximation of the
hydraulic fracture was developed which provided a reasonable approximation of well flow rate potential.
Subsequently, the individual well historical performance in the full field model was matched using regional
adjustments to the transmissibility and pore volume as well as the wellbore connection factor permeability. In
those cases where the hydraulic fracture treatment was significantly different than used in the preliminary work,
the effective wellbore radius was also adjusted to reflect the changed fracture length.
The gathering system pressure drops were matched using pipeline pressures obtained near the end of the
historical production period. Adjustments were made to the absolute roughness of the pipe segments to tune the

model predicted pressures to the actual measured values.


Prediction cases were then run using the history matched integrated model to evaluate the schedule of infill
drilling required to maintain deliverability. Wells were constrained by a maximum rate of 5 MMscf/d
(141 E+03 m3/d), since this was the functional limit of the wellsite line heater packages. The reservoir simulator
was then allowed to control the prediction via group rate targets. A prioritized drilling queue was set up in the
reservoir model and wells were added from the drilling queue as needed in order to meet the specified targets.
As a comparison to the integrated model, a prediction was made with the reservoir simulator alone using only
VFP tables to model tubinghead pressure. Figure 3, shows the difference in the well scheduling between two
identical prediction cases. The integrated model requires that wells be drilled sooner (in excess of 3 months in
some instances), although the number of wells required is similar in both cases. Given that individual well and tiein costs are on the order of 1.2 MM $CAD, this represents a substantial acceleration of capital expenditure
compared to the VFP run.
Application of Integrated Simulation Models
AEC had received notice from another participant in the Sexsmith plant that they intended to install field
compression at their wells located near the southern end of the Sexsmith pool. AEC was requested to participate
in the capital cost of the facility, since they also had wells nearby that could benefit from a reduction in wellhead
pressure. AEC management required a timely answer, but also wished to evaluate other alternatives such as line
looping and installing additional plant inlet compression. This is not as trivial a problem as it first seems, since all
wells, both current and future, will be impacted by the changes to flow rates, line pressures and potential leaseline
drainage. By using the integrated model, it was possible to evaluate all pertinent options and present
recommendations within a few days.
In another field case study, the integrated models were used to derive a development plan for the Maxhamish
Field located in a remote, winter access area near the British Columbia and Yukon border. This field had only a
few delineation wells drilled as of December, 1997. However, seismic and geologic control indicated a large
potential reserve base in an area of some 30 miles long by 8 miles wide. Initial runs were made with the models to
investigate the rate of depletion and the preliminary pipeline network design. Based on this work, approximately
25 more wells were drilled over the winter of 1997/98 and the results were incorporated into the model. At this
time, sufficient data existed so that the model could be run to finalize the depletion plan. Since the area is winter
access only, it is critical that sufficient wells are drilled and tied in each winter to offset the natural decline in
productivity over the rest of the producing year. The use of the linked models enabled well locations to be selected
that minimized the pressure drop in the low pressure (200 psi) gathering system and, at the same time, allowed the
transient deliverability decline to be modeled accurately. As an added benefit of linking the models, the sizing of
compression facilities required over the life of the field was also confirmed so that they could be ordered in time for
the winter of 1998/99, with project startup scheduled for April, 1999.

Recommended Future Enhancements


The greatest obstacle to the effective use of an integrated model has been found to be acknowledging and
understanding the significance of the large number of events that occur in both the reservoir and pipeline
simulation models. The integrated model allows the user to exercise a variety of automated controls such as
automatic drilling and workovers, bottomhole pressure limits, minimum, maximum and economic rates, high
water cut procedures and various equipment flow rate and pressure limitations. These can all cause wide
variations in expected production performance. Therefore, the objective of the simulation run must be discussed
beforehand with all parties involved so the events that occur during the simulation can be recognized and readily
understood.
To facilitate a better understanding of the control events that occur in both models, it would be advantageous
for these events to be summarized and exchanged so that both models could print them. It would also be useful for
the warnings issued from both models to be exchanged so that their influence on the simulation results can also be
acknowledged.
The examination of pipeline simulation results for large complicated gathering networks simulated over several
years can be very time consuming. This is especially true if both new wells and flowlines are added to the network,
compressor operating conditions are modified, or flow control or pressure control devices are installed. To better
understand the impact such changes impart on the pipeline simulation, one should carefully review such variables
as: flow rate and pressure changes; the velocity profile; the liquid holdup; and the changes in flow direction in the

network during each timestep. Similar to reservoir simulation models where fluid saturation changes can be post
processed and displayed, it would be useful to display the pipeline network showing the magnitude of such
variables as pressure drop, holdup, etc., according to a colour code scale. This would allow abnormal
circumstances that occur in the network to be more readily observed and detected.
Currently, only data critical to the simulation is exchanged through the interface between the models. The
addition of some data parameters such as well re-completion data to the information exchange would add further
functionality to the modeling software.
Enhancements to the reservoir model should include the addition of a linear flow well model to account for
short term but significant transient flow behavior, especially in tight gas reservoirs.

Conclusions
1. The movement by commercial software vendors to allow coupling of their models with other third party
software has created the opportunity for users to more rigorously model their reservoir and surface system than
ever before possible.
2. An integrated reservoir and pipeline simulation model is considered to generate a more realistic forecast of
production as compared to using a reservoir simulator relying on VFP tables or using separate, unlinked models in
parallel.
3. Constructing an integrated model brings the engineering resources of an organization together, with the model
acting as the focal point for the planning and management of the reservoir.

Acknowledgements
The authors thank AEC Oil and Gas Partnership for permission to publish this paper.

References
1. Startzman, R.A., Brummett, W.M., Ranney, J.C., Emanuel, A.S., Toronyi, R.M.: Computer Combines Offshore Facilities
And Reservoir Forecasts, Petroleum Engineer, May, 1977.
2. Emanuel, Alan, S., Ranney, Jon, C.: Studies of Offshore Reservoir With an Interfaced Reservoir/Piping Network
Simulator, Journal of Petroleum Technology, March, 1981.
3. Breaux, E,J., Monroe, S.A., Blank, S.S., Yarberry Jr., D.W., Al-Umran, S.A.: Application of Reservoir Simulator
Interfaced with a Surface Facility Network: A Case History, paper SPE 11479, June, 1985.
4. Stoisits, R.F., Batesole, E.C., Champion, J.H., Park, D.H.: Application of Nonlinear Adaptive Modeling of Rigorous
Representation of Production Facilities in Reservoir Simulation, paper SPE 24898 presented at the 1992 SPE Annual
Technical Conference and Exhibition, Washington, D.C. Oct 4-7.
5. ECLIPSE 100 Reference Manual, Schlumberger GeoQuest, 1997.
6. Hepguler, F., Barua, S., and Bard, W.: Integration of a Field Surface and Production Network With a Reservoir
Simulator, SPE Computer Applications, (June, 1997) 88.
7. FORGAS User Manual, Neotechnology Consultants Ltd., (1997).
8. Trick, M.D.: A Different Approach to Coupling a Reservoir Simulator with a Surface Facilities Model, paper SPE 40001,
presented at the 1998 SPE Gas Technology Symposium, Calgary, Alberta, Canada, March 15-18, 1998.
9. Open Eclipse Developers Kit, Schlumberger GeoQuest, (1997).

SI Metric Conversions
ft3
mile

x 2.817 399
x 1.609 344*

E-02 = m3
E+00 = km

MMscfd x 28.173 99
psi
x 6.894 757

E+03 = m3/d
E+00 = kPa

*Conversion factor is exact

Fig. 1Location of Sexsmith field in the Province of Alberta, Canada.

Fig. 2Sexsmith gas gathering network configuration.

Fig. 3Plot showing a comparison of the number of production wells


required to maintain a target production rate, using the integrated reservoir
and pipeline simulation model versus VFP tables.

Bibliography
Arvil C. Mogensen
Arvil Mogensen is president of ACM Consulting Ltd. located in Calgary, Alberta, Canada. He has been involved in a variety of oil
and gas consulting studies involving production and facilities and pipeline optimisation in Canada, USA, Middle East, Far East,
Australia, and Africa. He graduated from the British Columbia Institute of Technology with a diploma in Natural Gas and
Petroleum Technology, and graduated from the University of Alberta, with a BS in Civil Engineering.

Gregory R. Caswell
Graduated with a Diploma of Petroleum Technology (Reservoir) from SAIT in 1971. Has been involved in reservoir engineering
applications and reservoir simulation since graduation, both with independent exploration companies and major international
consulting organizations. Experience has focused on EOR, well test analysis and hydraulic fracture simulation and design. Formed
Caswell Petroleum Consultants in 1990 and offered services in reservoir engineering, well testing and analysis, drilling, completions
and facilities construction, mainly in the Western Canadian basin. Joined AEC Oil & Gas Partnership in 1997 as Technical Advisor,
Reservoir Engineering for the AEC West business unit. Author of numerous technical studies and co-author of three additional
published papers.

Bibliographies
Arvil Mogensen is president of ACM Consulting Ltd. located in Calgary, Alberta, Canada. He has performed and managed various
consulting studies involving oil and gas production, surface process facility, and pipeline optimization studies in Canada, USA,
Middle East, Australia, Africa, and South America. He holds a Diploma in Natural Gas and Petroleum Technology from the British
Columbia Institute of Technology, and a BS degree in Civil Engineering from the University of Alberta.

You might also like