Linking Reservoir and Surface Simulators How To Improve The Coupled Solutions Barroux2000

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

SPE 65159

Linking reservoir and surface simulators: how to improve the coupled solutions
C. C. Barroux, Institut Francais du Petrole, P. Duchet-Suchaux, TotalFinaElf S.A., P. Samier, TotalFinaElf S.A., R. Nabil,
Gaz de France

Copyright 2000, Society of Petroleum Engineers Inc.


coupled way the reservoir model and the surface model, for
This paper was prepared for presentation at the SPE European Petroleum Conference held in enabling time savings (by avoiding multiple information
Paris, France, 24–25 October 2000.
returns between reservoir and surface engineers), and for an
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
improved reliability of the global solution.
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to Since the pioneer work of Dempsey1 in 1971, the
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 development of a coupled resolution of reservoir and surface
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
models, firstly for application in gas field development and
for commercial purposes without the written consent of the Society of Petroleum Engineers is production optimization, has been extended some years after
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 by Chevron for oil field developments2-4 in offshore area.
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O. Since then, several other major companies have developed
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.
their proprietary integrated reservoir/surface solution5-8, and/or
have integrated their proprietary reservoir simulator with a
Abstract commercial network software9,10. To our knowledge, the only
A full analysis of the problem of coupling two independent reported experience of a compositional integration is from
industrial simulators, a reservoir simulator and a steady-state Litvak11-13. In these integrated solutions, the two programs are
network simulator, has been performed. A classification of the merged in a unique one (solution named ‘fully integrated
industrial needs has been obtained from the examination of solution’ in what follows) ; such a procedure has also been
several actual cases studies (fields at development, or used by Trick14 for integrating two commercial software
production stage, requiring performance optimization through Forgas and Imex.
short to long term forecasts). Several coupling configurations Another integration method, enabled by the emergence of
and specific requirements, for both reservoir and surface message passing libraries as PVM (Parallel Virtual Machine),
software, have been identified to satisfy the industrial needs. is to couple independent programs through an interface. In the
Various commercial solutions have been screened to evaluate coupling development of two commercial simulators, Eclipse
the adequacy between the software functional capacities and as reservoir simulator, with Netopt (Hepguler15) or with
the identified configurations and requirements. Limitations Forgas (Trick16), the interface which controls the software
have been brought to evidence. communication has been implemented in the surface
The paper, issued from the work of a multidisciplinary simulator, the reservoir simulator being a slave process. The
team, presents the details of each step described briefly here coupling location, retained for linking two independent
above. Practical solutions for improving coupled solutions, i.e. simulators, is either at the bottom hole (Trick), or at the
to go beyond the identified limitations and to reduce the reservoir level with an Inflow Performance Relationship (IPR)
elapsed times of the coupled simulations, are provided. An represented in both reservoir and surface models (Hepguler).
improved mutual understanding of the needs of each discipline Apart from the integration procedure and the coupling
concerned by the integration of reservoir and surface has been location (surface, bottom hole or reservoir level), the
experienced as essential for a successful integrated approach published solutions differ from each other in particular about
of field optimization. It is intended in the paper to contribute the nature (explicit or implicit) of the algorithm, the
to a better communication between the various upstream equilibration frequency (the frequency for calling the surface
petroleum skills: reservoir, production, surface and process network resolution), the type of fixed boundary conditions
engineers. exchanged between the programs. For all these solutions, a
Introduction strong concern is minimizing the computation times. For the
It has been well recognized that coupling surface network and Eclipse/Forgas link, significant savings in computation time (a
reservoir simulations is a critical technology for field factor of 30) have been achieved using a coupling at the
development/optimization studies, particularly in the deep Newton iteration level compared to a coupling at the time step
offshore area, or when many wells share the same surface level as used by Hepguler. For fully integrated solutions, more
network. Various methods have been proposed to solve in a
2 C.C. BARROUX, P. DUCHET-SUCHAUX, P. SAMIER, R. NABIL SPE 65159

fundamental techniques for accelerating implicit resolution test 3. For test 4, the communication, between the Unix
methods have also been proposed, among which one finds platform used for the reservoir simulation and the PC platform
domain overlapping17, domain decomposition18, and solution used for the surface simulation, has been established via an
preconditioning19. Ethernet network and PVM.
Up-to-now, a full integration appears to be still often The test production results, not shown here, are quite
preferred to a coupling through an interface software. similar. Two observations can be made: 1) coupling two
However, the coupling through an interface could offer more independent simulators through an interface does not result
flexibility in software choices when coupling two independent necessarily in a worse CPU time than that obtained with a
simulators, commercially available or not, provided that each fully integrated solution, 2) communication between different
simulator has to be "opened", i.e. it can exchange data with an platforms through a computer network is a factor for
external software during the simulation. Crossing the increasing elapsed simulation times. It is worth mentioning
information got from the literature, and from the screening of that, if here the time spent in the surface simulation is
various software, have allowed for getting a better relatively small compared to the time spent in the reservoir
understanding of this situation and for providing tracks on simulation, it is because of the case simplicity (a few wells, a
potential ways for improving the interfaced solutions. simple surface network, an essentially single phase flow,
Industrial needs from examination of various cases Black-Oil modeling).
Two case studies presented hereafter are used to discuss Case 2
particular results obtained with various coupled or fully To optimize the plan of development of a deep offshore oil
integrated solutions. Two other case studies are described to field with gas lift, simulations have been performed using the
focus the attention on needs not yet well satisfied by coupling of two independent software through an interface.
commercial solutions. They allow also for illustrating some Thirteen wells are producing through a rather complex
contexts where a reservoir/surface integrated approach is network including seven trunk lines. One of the simulation
needed. Since the objective is not to promote any software, cases has been used to test two frequencies of equilibration
and also because of the rapid software evolution, possibly between the reservoir simulator and the network simulator: at
invalidating in a next future any to-day ranking of software each reservoir simulation time step, at each month. Figure 1
quality, it has been preferred to avoid any mention of a trade shows that the run time (as a function of the simulated time) is,
mark. for an equilibration at each time step, about 50% higher than
Case 1 for an equilibration at each month.
The aim of the study is to explore the possibilities of
putting into production a high pressure gas reservoir through Fig. 1 – Influence of equilibration frequency on computing time
an existing surface network gathering the production of
Computing Time (hours)

16
several gas reservoirs already largely produced and depleted. End of the
Date of compression, compression sizing together with 12 production
optimized production scenario are expected from the study. plateau
From a coupling point of view, the case is rather simple since 8
only a few number of wells (coupling points) are involved, and Reservoir
since the flow is essentially single phase flow. Specific simulation
4
requirements for the surface network simulator are an alone
optimization functionality and an option to represent some of 0
the reservoirs as tank models. Table 1 gives the running 0 0.5 1 1.5 2
computation times (RT) of four tests with different coupled Sim ulated Tim e (Days x 10**3)
and fully integrated software, normalized with respect to a Equilibration frequency : Time Step Month
reference running time. The reference running time (RTref) is
the time obtained for the reservoir simulation alone (without Fig. 2 – Influence of equilibration frequency on simulation results
any coupling to a network simulator), run on a Unix platform.
Qo
Table 1: test simulation conditions
Production Rate

Test Nature of RT/RTref Simulation Platform


# coupling Reservoir Surface
1 Fully integrated 1.1 Unix Unix
2 Fully integrated 1.7 Unix Unix Qw
3 Through interface 1.2 Unix Unix
4 Through interface 2.5 Unix PC
0
For tests 1 to 3, the reservoir simulation and the surface
0 0.5 1 1.5 2
simulation have been processed on a unique Unix platform,
the same as that used for the reservoir simulation alone. PVM Simulated Time (Days x 10**3)
has been used for communication between the simulators in Equilibration frequency : Time Step Month
SPE 65159 LINKING RESERVOIR AND SURFACE SIMULATORS: HOW TO IMPROVE THE COUPLED SOLUTIONS 3

The results show also that the simulation is faster in the approach has been recognized as essential. A first step
production plateau period than in the decline period, the time underway is building a model for the actual complex network.
savings appearing larger in the decline period than in the A second step will be to perform coupled reservoir/surface
plateau period. Figure 2 illustrates that the results of oil and simulation studies on a few last years of production to verify
water field production rates obtained during five years of the adequacy between the reservoir model and the surface
production with both equilibration frequencies are quite model in a context of history match. It is wished, for the first
similar, this demonstrating that a coupling at each time step is history match simulations, to use a loose coupling, i.e. to
not optimal in this case. couple explicitly the simulators at the well head level, the
Both reservoir and surface models are Black-Oil (BO). The reservoir model passing rates as boundary conditions to the
run time, even with the reduced equilibration frequency, is network model. To keep consistency with the rate control by
much larger (about ten times) than the time needed by the the reservoir model, it is also wished that the network
reservoir simulation alone. For a coupled compositional simulator works under the reservoir simulator control. These
simulation, potentially requested in case of limiting options, not developed in the available commercial interfaces,
composition specifications and/or for providing compositions are discussed in the following sections.
to surface process simulations, computation time should have Comments
been even larger. In fact, we have found no commercial A common point in the cases briefly presented above is the
solution for compositional coupling of two independent very large time scale of the simulations compared to the
software. To our knowledge, the only reported compositional typical time scale of transient phenomena occurring in wells,
experience is from Litvak12 and it seems to correspond to a risers or at the surface level. The transient effects are then not
fully integrated solution. Besides, although it could have been considered in what follows, although, in the deep offshore area
of interest to model the gas re-injection network for gas lift, no where wells are complex and in a smaller number than on
commercial solution allowing for a simultaneous resolution of shore, it could be envisaged to get more accuracy by coupling
production and injection networks has yet been found. the reservoir simulator to transient well and network
Case 3 simulators.
A gas field, composed of hundreds of non communicating For the present discussion, a well (or a branch in the
traps, is at the development stage. Most of the eighty strings network) is considered, as usual in the reservoir engineering,
are planned to produce several traps in commingle. Trap either open to flow or closed, and only steady-state (or pseudo
depths range between 2000 and 4000 m, varying fluid steady-state) flow has to be modeled in the network simulator.
compositional trends having been identified. Interval between As a summary of the industrial needs appeared from the
top and bottom perforated traps in a given well can reach 1500 cases presented here above, and to allow the user for choosing
m. Cross-flow between traps produced by a same well is the operating mode, what is required is:
possible during the exploitation. Because of rapid trap pressure For both simulators:
decline and surface network limitations, a reservoir/surface - Fast resolution ;
integrated approach has been necessary to optimize the - Portability ;
development plan (production plateau, well drilling schedule, - Extended Black-Oil (in particular with variable gas
production facilities, ..), and among other tasks, to interpret the gravity and possibility for compositional tracking), and
early production stage in terms of pressure and composition in compositional thermodynamics ;
the individual traps. Production is collected through a rather - Possibility of passing pressure, rates, and compositional
simple network. To couple a reservoir model to a network information at well head level, at reservoir level, at
model at the trap level, a specific requirement for the network bottom hole level ;
model is that it should be able to manage a situation where - Controllability from an external process.
each common boundary point may change from a source status For the interface:
to a sink status (or vice versa) during the simulation without - Robustness of the coupling scheme ;
any a priori knowledge on the dates at which the changes - Optimization functionality (see application examples21,22);
occur. Although an in-house software (approach similar to that - User defined and/or intelligent equilibration frequency ;
described by Puchyr20) exists, we have looked for alternate - Allowing for BO or compositional coupling ;
solutions from the to-day market. So far, no commercial - Option for boundary conditions: rates or pressure.
solution well suited to the problem size and specificity has Specific requirements for the surface simulator, in addition
been found. to single or multiphase flow modeling and IPR handling, are
Case 4 flow reversibility at coupling location, the ability to represent
A large oil field (about one thousand wells) has been oil production together with gas or water re-injection
produced for more than thirty years by gas injection, water networks, and a card time functionality for entering the
injection and gas lift. It is planned to double the production in network modifications during the whole simulated time.
the next twenty years. Up to now, the reservoir simulations, Operating modes
and in particular the history matching, have been run Already introduced in former papers (Breaux4, Puchyr20), the
'conventionally', i.e. without an integrated reservoir/surface distinction of two operating modes, ‘history matching’ and
approach, but for the further field developments, such an ‘prediction’, interesting from a general practical point of view,
4 C.C. BARROUX, P. DUCHET-SUCHAUX, P. SAMIER, R. NABIL SPE 65159

should also be considered when coupling reservoir and Fig. 4 - Configuration 2: coupling at reservoir level with IPR
overlap
surface. Indeed, in a history matching context, one knows
which wells are opened, and well rates can be estimated from Network Model
well production tests or from a rate allocation procedure:
degrees of freedom are less than in a purely predictive context,
where well performance, dates for well opening and shutting
are estimated during the simulation using some production J2 J2
optimization process at the surface level. As, in a predictive
Reservoir Model
context, rates are unknowns, reservoir pressure is an
appropriate (relatively slowly varying) boundary condition for IPR1
the network simulator, and a coupling location at the reservoir J1

level is advantageous ; but, for history matching, especially for


roughcast simulations, coupling at the well head level is more S1 : Reservoir cell
IPR1
practical. It is noticeable that, as reported by Tingas10, two J1 : Bottom Hole
J2 : Well Head
J1

solutions are available in the GCOMP/PIPEPHASE S1 S1


integration: the coupling location can be at the well head or at
the bottom level ; coupling at the well head, although less Fig. 5 - Configuration 3: coupling location at reservoir level with
accurate in wellbore hydraulics calculations than the (well + IPR) overlap
alternative solution, has been found adequate in many cases,
S1 : Reservoir cell
and faster. Although no reason other than the solution fastness J1 : Bottom Hole
is provided to sustain the development of the two solutions, J2 : Well Head
Reservoir
Tingas contribution gives some support to our analysis. model
A last point concerning the incidence of the operating Network model
J2
mode on a coupling through an interface, is that, if it comes
naturally the reservoir simulator to be a slave process for
forecasting (the production logic is optimized at the surface
level), it should be practical, as expressed in case 4 J1 J2
IPR1
description, to reverse the situation for history matching.
Furthermore, as history match is usually followed by forecasts,
IPR1
a switch option between the two operating modes should also S1
J1
be practical. The commercial interfaced solutions offer
S1
potential flexibility in the reservoir simulator choice, but are
less flexible concerning the coupling location, and concerning Fig. 6 - Configuration 4: coupling at trap level with IPR overlap
the control of the surface model from the reservoir simulator
S1, S2 : Trap Network model
or from a third-party interface. J1, J2 : Bottom Hole
Four coupling configurations are discussed in next sections J3 : Well Head
(Figures 3 to 6). The pressure drops in the well tubing are
represented, for configurations 1 and 2, respectively in the J3
J3
reservoir model and in the surface model, while they are
represented in both models in configuration 3. Reservoir model
S2 S2
Fig. 3 - Configuration 1: coupling at the well head level IPR2 IPR2
J2 J2

Network Model

IPR1 IPR1
J1 J1

S1 S1
Reservoir Model

J2
J2 Configuration 3 may then allow for switching from a
coupling at the well head, usable in a history match mode, to a
coupling at the reservoir level, more appropriate in a
S1 : Reservoir cell predictive context. Figure 6 pictures an extension of
J1 : Bottom Hole
J2 : Well Head configuration 2 concept, denominated as configuration 4, for
representing wells producing several individual traps (as
IPR1 required in case 3).
IPR1
J1
J1
These configurations are discussed after a presentation of
S1 software functionality involved in the coupling procedures.
S1
SPE 65159 LINKING RESERVOIR AND SURFACE SIMULATORS: HOW TO IMPROVE THE COUPLED SOLUTIONS 5

About network simulators The procedures used for coupling independent


Advanced steady-state network simulators handle single or simulators15,16 are limited to a Black-Oil fluid representation.
multiphase flow, offer a card time functionality, have various Extension of coupling to other thermodynamic contexts is
compositional options including fully compositional with discussed in a separate section.
equation of state. They usually accept rate or pressure as In the coupling algorithm described by Trick16, the surface
boundary conditions at source and sink nodes. Some of them simulator uses the flow rates computed by the reservoir
are portable. simulator to compute the pressures throughout the surface
Other features have to be emphasized in the context of network and the well bottom hole pressures at a predefined
reservoir/surface integration: datum depth (BHP). The BHP computed values are passed to
1. The pressures and temperatures in the network are usually the reservoir simulator which uses them to revise the flow
solved in a coupled way; rates during the next Newton iteration. The convergence is
2. The limitation for the flow direction at the source nodes obtained when the difference between flow rates from one
(precluding the representation of cross-flow in wells) iteration to the other is less than a user defined tolerance. To
comes from a distinction of the boundary nodes, as source speed the convergence, the surface model passes the average
nodes, - where, in addition to a pressure (or rate) of BHPs from the previous iteration and the current iteration to
condition, a compositional information and a temperature the reservoir simulator. In case of rate oscillations, the BHP
have to be fixed -, and sink nodes where only a pressure may be determined from computing the intersection of the
or rate condition has to be fixed ; inflow performance and tubing performance curves.
3. Another limitation concerns separator representation: In the model described by Hepguler15, there is, as for
usually, only one output of the separator can be connected Trick, a single coupling point per well, but the coupling
to a downstream network (chosen as the oil output for location is at the reservoir level, i.e. that the first pressure drop
representing an oil production network) ; it follows that a law at a source node shared with the reservoir simulator
separator has not a node status, but is considered as a corresponds to an Inflow Performance Relationship (IPR).
device in a link between two nodes, and this precludes the Although several types of IPR are implemented in the
possibility of simulating gas or water re-injection. surface simulator as in the reservoir simulator, the IPR, used in
To understand the origin of these limitations, we have the reservoir/surface coupling context, is a linear IPR for each
enumerated, for the network problem, the number of phase p, global for the well (not for each perforation):
equations, the number of unknowns, the difference
Qp = Ap − Bp BHP ........................................................ (1)
corresponding to the number of boundary conditions necessary
to the problem closure. A simple description is provided in which is handled in the surface simulator as:
appendix A. It appears that, in a steady-state network model :
Qp = PIp (PB − BHP) , with PB = Ap/Bp , and PIp = Bp (2)
- the fact of forcing a rate or a pressure condition at each
boundary node precludes a description, in the network For a coupling location at the reservoir level for a given
model, of all separator downstream branches, well, the reservoir simulator has then to pass the two
- the fact of forbidding an overwriting of the temperature coefficients per phase to the network simulator. The ratio
and composition conditions at a source node precludes the Ap/Bp, used as a pressure boundary condition, and Bp
modeling of a reverse flow at a source node, then coefficients are kept constant during the resolution of the
precludes to represent the cross-flow in wells. network problem, while the pressure BHP is recalculated. If
Overcoming these limitations is possible by rewriting the flowing bottom hole pressures calculated by the two
and/or rearranging the system of equations, but may require an simulators are not within a predefined tolerance, the well rates
heavy software development. calculated by the surface simulator are passed back in a
A more easily tractable point of improvement, which consistent form to the reservoir simulator as new targets to
should allow for diminishing the computation time, concerns redo its time step calculations.
the temperature calculation: as long as the rate variations in What is not defined precisely in these two former
the network are not considerable, it is unnecessary to contributions is the type of conservation law used to ensure the
recalculate the temperatures. Disconnecting pressure and consistency of flow rates when passed from one simulator to
temperature calculations, and the use of a smart rate criteria the other. Volume conservation per phase at bottom hole
for network temperature update, should result in appreciable conditions is physically meaningful only if the two simulators
computation time savings. reproduce exactly the same phase thermodynamic behavior. In
About interfaces a Black-Oil context, a common situation is that fluid
Presently available interfaces for linking reservoir and network properties are defined in tables as functions of pressure in the
simulators are integrated in the network software as an input file of the reservoir model while they are calculated from
additional (possibly optional) module. An optimization internal analytical correlation in the network simulator. In such
functionality is usually provided in the interface package. a situation, passing volumetric rates at bottom hole conditions
Portability and user-defined equilibration frequency were not from surface to reservoir or vice-versa is not accurate and may
met so far in a same commercial software. cause trouble to obtain the convergence. What is the most
appropriate in a Black-Oil context is to exchange phase
6 C.C. BARROUX, P. DUCHET-SUCHAUX, P. SAMIER, R. NABIL SPE 65159

volumetric rates at standard conditions since they are Well boundary conditions and working point
equivalent to component (oil, gas, water) molar rates. In what determination
follows, whatever the thermodynamic context, it is assumed Per well in a reservoir simulator, at least one limiting
that the data exchange conserves the number of moles per pressure condition PLIM (minimum limit for producers,
component, a mass conservative law being the only acceptable maximum limit for injectors) , either at the well head or at the
alternative. bottom hole, and one maximum rate condition (QMAX:
In the Hepguler description, a first underlying assumption volumetric rate for a specified phase or group of phases),
is that the ratio Ap/Bp is the same for all phases. A second either in standard or bottom conditions, have to be fixed by the
implication is that each Ap and Bp would not depend on BHP. user. When several pressure constraints are entered for a same
If the IPR coefficients depend on BHP, obviously the well, for example one at the bottom level as a mechanical
coefficients have to be updated by the reservoir simulator at failure threshold, and one at the well head level, the simulator
each iteration needed to reach the convergence at a particular retains the most constraining condition. The usual procedure
time step, and the convergence between reservoir and network consists in first calculating well rate using PLIM as boundary
solutions may be more difficult to obtain. condition. If the calculated well rate does not exceed QMAX,
Procedures for generating linear IPR coefficients are the effective constraint is PLIM ; in the opposite case, the
described in appendix B. It is shown that linear IPR, at the QMAX constraint applies.
well level, and written for molar fluxes, is usable even when When the limiting pressure condition is set at well head,
several perforated grid blocks contribute to the well the pressure drop in the well between the well head and a
production. However, when sophisticated options for wellbore predefined well datum depth (Zref) has to be represented in
calculations (as implicit fluid densities) are activated in the the reservoir simulator, usually through hydraulic look-up
reservoir model, the IPR coefficients should depend directly tables. If the oil phase is the reference phase, these tables
on BHP, and a reservoir/surface coupling at each Newton reflect the relationship between the flowing bottom hole
iteration in the reservoir simulation becomes necessary. pressure (BHP) at Zref, the well head pressure (WHP), the oil
In the convergence procedure used by Hepguler, if the flow phase rate (standard conditions), for various values of gas-oil
rates and flowing bottom hole pressures calculated by the two ratio and water-cut.
simulators are not within a predefined tolerance, the well rates The determination of well working conditions (or working
calculated by the surface simulator are passed back in a point determination) is done only at the beginning of the time
consistent form, for the next iteration, to the reservoir step in the standard situation where the bore hole pressure
simulator as new targets to redo its calculations. If there is a gradients (from well datum depth to each active perforation)
single active perforation in the well, component molar rates are hydrostatic and handled explicitly: the fluid densities used
can be introduced directly in the material balance equations of to calculated the pressure gradients below the datum depth are
the perforated grid block. However, if there are several active estimated using the well flowing pressures of the preceding
perforations, the reservoir simulator has then to allocate the time step. The error induced on the flowing pressures in the
well rate passed by the surface simulator to the various active well by a bad PLIM estimate is corrected in the further time
perforations. In this latter case, it appears, from what is steps.
discussed in appendix B, that, when standard wellbore Obviously, the use of hydrostatic pressure gradients below
calculations (explicit fluid densities for calculating hydrostatic the well datum depth becomes questionable in cases similar to
pressure gradients in the well) are used, a better choice should case 3, this explaining why a coupling at the trap level is
be to pass to the reservoir simulator the bottom hole pressure envisaged in this particular case.
calculated by the surface simulator instead of the rates. This As already stated by Startzman2, bottom hole pressure are
would require a small development in the reservoir simulator. hard to forecast with accuracy. It is why it is very often
About reservoir simulators preferred to set limiting pressure condition PLIM at well head.
Advanced reservoir simulators offer most of the features listed If a producer effective constraint is QMAX, the reservoir
before except that most of the reservoir software have no open simulator computes a well head pressure (WHP) higher than
functionality (for data exchange during the simulation and PLIM. It means that some kind of well rate regulator is
controllability from an external process). In addition of the implicitly represented in the reservoir simulator, the upstream
work required for opening the simulator, some other specific pressure of which being WHP, and its downstream pressure
developments presented hereafter should be necessary for being PLIM: this regulating device corresponds to a simplified
linking the reservoir simulator with the network simulator at representation of a choke, stating that it exists a choke opening
the well head level and at the trap level. for which QMAX rate is feasible when device upstream and
Given the importance of well pressure and rate constraints downstream pressures correspond respectively to WHP and
when linking reservoir and surface simulators, a special PLIM. PLIM should then correspond to the pressure (PDOR)
attention is first devoted to remind how the wells work in a calculated downstream a regulator, put at the well head in a
reservoir simulator. Other aspects concerning the fluid network simulation. Although not rigorous, this approach is of
representation, and methods for delumping the results of particular interest when used in reservoir/surface coupling for
reservoir simulations (compositional and Black-Oil), are reducing computing time through a partial decoupling of the
discussed in a separate section. simulations, as requested in case 4.
SPE 65159 LINKING RESERVOIR AND SURFACE SIMULATORS: HOW TO IMPROVE THE COUPLED SOLUTIONS 7

Special requirements for coupling at the well head Coupling configurations


For a link at the well head level, a temperature has to be For determining the range of application of the configurations
assigned at the well head since a temperature is one of the described hereafter, a simple statement is that the simulator
boundary conditions requested at each source point by the which fixes the well rate condition to the other simulator
network simulator. The well head temperature is not explicitly handles also the rate allocation strategy. Rate allocation
used in a standard reservoir simulator. However, the well head strategy is typically at the surface level in a predictive context
temperature is usually one of the outputs of the preprocessing (using some kind of optimization process external to the
allowing for the generation of look-up tables which are the reservoir simulator), at the reservoir level in an ideal history
standard way, in a reservoir simulator, to handle multiphase match context (deterministic approach, assuming known well
flow in the well bore. Then, the well head temperatures can be rates). The discussion holds for field production modeling,
provided as a supplementary column in the look-up tables. except when injection problems are mentioned explicitly. In
With the implementation of an appropriate interpolation law, it the reservoir model, it is assumed that the computations of
will allow the reservoir simulator for passing the well head fluid densities below the well datum are explicit. A coupling at
temperature to the network simulator. the level of Newton iterations in the reservoir simulation (way
Special requirements for coupling at the trap level for accelerating the convergence16), although possible, is not
In the situation where a given well produces several traps discussed here.
(not connected) in commingle, linking the network simulator Configuration 1: coupling at the well head level
with the reservoir simulator at the trap level means that: Wellbore hydraulics are modeled in the reservoir
- Well tubing hydraulics are not represented in the reservoir simulation only.
model ; In the surface model, each well head is a source point. If no
- Each coupling point in the well has to be identified ; only rate control is represented downstream the source point, the
one coupling point per trap is envisaged, even if a trap is rate calculated by the reservoir simulator is the boundary
connected to the well through more than one perforation. condition fixed to the source point in the surface simulation,
- A datum depth has to be defined for each trap in the well ; the well head pressure from the reservoir simulator being
- The bottom hole pressure at each trap datum depth in the possibly used as a first estimate by the surface simulator. For a
well has to be calculated ; producing well, the well head temperature (see the paragraph
- The reservoir simulator will have to calculate IPR 'Special requirements for coupling at the trap level') and fluid
coefficients for each active trap in the well, and to pass composition have also to be fixed. The rates being not
them to the network simulator ; revisable by the surface simulator, the production strategy is
- Pressure constraints have to be implemented at the level introduced in the reservoir model.
of each trap ; The rates passed to the network simulator can be those
- If a given trap is connected to the well through more than calculated at the beginning of the time step from the well
one perforation, the trap molar rate (per component) has working point determination in the reservoir simulator.
to be computed as the sum of the molar rates of all active The network model is solved using the specified rates. The
perforated grid blocks belonging to trap ‘t’. pressures recalculated at the source points correspond
A way to fulfill all these requirements without reservoir physically to the well PLIM constraints in the reservoir model
code modification is to declare for a given actual well as many and are named (PDOR) even if the regulator is not explicitly
independent pseudo wells in the reservoir model as the number represented in the surface model. If the pressures (PDOR) are
of perforated traps in the well. Adoption of such a procedure less, within a user defined tolerance, than the well head
in case 3 should result in the introduction of more than 1000 pressures (WHP) calculated by the reservoir model, the
pseudo wells to represent the 80 actual strings. reservoir rates can be considered as feasible. In the opposite
If traps are produced in commingle, it can occur that the case, the pressures (PDOR) can be passed to the reservoir
type (producer, injector) of some pseudo wells has to be simulator to overwrite the PLIM constraints and redo the well
switched from one type to the other, to enable cross-flow working point calculations, this step being repeated if needed.
between traps. This means that the pseudo wells, Once consistency is obtained, the reservoir simulator can
corresponding to traps potentially in a cross-flow situation, proceed to the time step calculation. If the rate target of a
have to be declared twice, as producer and injector. The given well cannot be sustained for the time step, the well rate
introduction of a non standard well type (with dynamically will be computed using the well PLIM constraint. It follows
reversible well rate) in the reservoir simulator is necessary for that overwriting the PLIM constraint with the PDOR value and
avoiding the pseudo well duplication. recalculate the well working point may be advantageous even
Options for assigning one constant temperature par trap, when the rates are found feasible from the first call to the
one initial gas gravity per trap (in a Black-Oil context) should surface simulator. Recalculating the time step is advisable if
be offered to consider temperature and compositional the rates at the end of the time step differ notably from the
gradients usually encountered when the range of trap depths is rates at the beginning. Cutting the time step may be indicated
large. In addition, some kind of compositional tracking should in case of strong rate decline.
be useful in particular to follow for gas gravity changes due to The interesting feature is that, as long as PDOR (from the
cross-flow between traps. network simulator) are less than WHP (from the reservoir
8 C.C. BARROUX, P. DUCHET-SUCHAUX, P. SAMIER, R. NABIL SPE 65159

simulator), surface and reservoir simulations can be reservoir model rate value by the regulator. At the end of
considered as consistent with no need of further iteration. the surface simulator computations, the recalculated BHP
Passing PDOR from the surface simulator to the reservoir are passed to the reservoir simulator.
simulator to overwrite the PLIM constraint of next time steps d) The reservoir simulator recalculates the well working
can be used for slacking the coupling equilibration frequency, points (using updated BHP), and proceeds to the time step
noting that a compromise has always to be found between calculations.
speed, convergence robustness, and solution accuracy. e) Consistency between the rates obtained in steps c) and d)
Such an approach is close to that of Chevron3,4. The above is checked. If the rates are consistent (within a predefined
description is certainly not optimal during the well decline tolerance), one can proceed to the next time step. In the
period. Computing time savings and rate oscillation remedy opposite case, calculations are redone from step b).
may be obtained using anticipation of the rate and pressure Such a procedure could be preferable to that described by
evolution from the preceding time steps3,5 (or preceding Hepguler15, for saving computation time in case of multiple
iterations16) for improving the variable updates exchanged by perforations per well. Whatever the procedure used, the rate
the simulators. allocation strategy is manageable from the surface simulation
Configuration 2: coupling at reservoir level with IPR (of particular interest in a predictive context). Drawbacks of
overlap the configuration are the difficulty for entering reasonable
Wellbore hydraulics have to be represented in the surface bottom hole PLIM estimates in the reservoir model input file,
model. As IPR is represented in both reservoir and surface and reduced flexibility for slacking the equilibration
domains, IPR coefficients are part of the data exchange. In frequency. Indeed, forcing, in the reservoir simulation during
standard reservoir simulations, grid cell temperatures are the periods where the surface simulator is not coupled, the
constant. The temperatures at the coupling points can then be wells to work at bottom PLIM supposes the presence of a
fixed in the input file of the surface model, temperature data bottom hole rate control equipment wells. As it is usually not
exchanges being then unnecessary. the case, the only remaining possibility for controlling the
Methods for deriving the coefficients (formulated for wells in the reservoir simulation is to force the wells to work
molar flux), when there are several active perforations for a at constant rate, this situation being potentially troublesome in
given well, is given in appendix B. It is shown that, rather than the rate decline period.
exchanging the IPR coefficients, an alternate solution Configuration 3: coupling location at reservoir level with
(allowing for compacting the information) consists in passing (well + IPR) overlap
one global IPR coefficient (productivity index), one pseudo This configuration uses an overlapping domain between
'block pressure' (equivalent to the ratio A/B), and molar the surface and the subsurface models extended from the IPR
fractions describing the global fluid composition, these to the well head. A necessary condition is that well hydraulics
quantities being obtained by a rearrangement of the IPR are handled in the same way in both simulations, for example
equations. by introducing in the surface simulator the possibility of
Here (see Figure 4), a rate control device (simply representing well hydraulics through the same flow tables as
reproducing a QMAX logic similar to that used for well used by the reservoir simulator (flow tables being
control in reservoir simulators) is present downstream the well advantageous for saving computation time). The flow tables
head in the surface model. A possible algorithm is: have then to include a column for the well head temperature
a) At the beginning of the time step, the reservoir simulator (see par. 'Configuration 1').
determines the well working points (computation of Assuming PLIM constraints fixed at the well head in the
component molar rates at each active perforation) using in reservoir simulation, the proposed computation algorithm uses
particular the QMAX and PLIM constraints and the the same steps a) and b) as in the preceding section. In step c),
flowing pressures from the preceding step to update the pressures computed downstream the regulator by the surface
wellbore fluid densities. simulator are also passed to the reservoir simulator (to
b) The IPR parameters (A, B) are computed. The well overwrite the PLIM condition). If the maximum rate values
reference rate, fluid composition details, IPR parameters used at the regulator level have been revised by the surface
and BHP are passed to the surface simulator. simulator for optimizing the rate allocation, these values have
c) Maximum rate values assigned to the control devices may also to be sent (to overwrite the QMAX constraint). In step d)
be either the calculated well rates from step b) or from the well working points are determined using the updated PLIM
surface simulation (input data file, optimization process, (and QMAX) constraints. BHP sent by the surface simulator
..). The principle of surface simulator calculations is first can be used for processing the wellbore fluid densities needed
to compute the well rates using the pseudo block in the computation of individual perforation rates. Note that, if
pressures as boundary conditions at the source nodes, then a well switches from a QMAX to a PLIM constraint during the
to look for honoring the regulator maximum rate for any time step calculations, the use of the PLIM condition updated
well the rate of which is found higher than its maximum from step c) is profitable to the solution consistency. Step e) is
rate constraint. The IPR parameters being the same in unchanged. Extending the PLIM updates to the next time steps
both simulations, the computed bottom hole pressures of a may allow for loosening the equilibration frequency in a
given well are equal if the rate is constrained at the reasonable way.
SPE 65159 LINKING RESERVOIR AND SURFACE SIMULATORS: HOW TO IMPROVE THE COUPLED SOLUTIONS 9

Such a configuration offers then the flexibility of choosing [F] being a very detailed fluid representation used as a
tight and/or loose coupling of independent simulators and is reference to build [R] and [S].
adequate in both history match and prediction contexts. In the second sequence, an efficient method for delumping
Configuration 4: coupling at trap level with IPR overlap the surface simulation should be provided, but from our
With the procedure proposed in the paragraph 'Special knowledge no method has yet been proposed. If the fluid
requirements for coupling at the trap level', configuration 4 representation [S] is more detailed than the [R] representation,
becomes a particular case of configuration 2. the remedial consists in delumping [R] representation directly
The very specific requirement is the reversibility of flow in the [S] representation.
sign at coupling points, for a switch from a producer status to Another difficulty may arise from the exchange of phase
an injector status and vice-versa, some developments being rates, either to be used in the convergence algorithm, or as a
needed in both reservoir and surface simulators as discussed in simulation QMAX type constraint. If both simulations are
former sections and in the appendices. Main modifications in compositional, phase rates result from a flash operation using
the step c) of the procedure described for the configuration 2 typically an equation of state. Results of flash operations at
are: firstly, maximum rate values to assign to the well head given pressure and temperature, from thermodynamic
regulators are managed by the surface simulator only, packages developed by independent parties, even using a same
secondly the surface simulator has to pass the fluid EOS, components and compositions, are usually not rigorously
composition information to the pseudo wells corresponding to equal. To avoid convergence difficulties, a remedial can
traps in a cross-flow situation. consist in converting, before the exchange between the
Fluid representation simulators, any phase rate in component molar rates, each
Black-Oil (BO) fluid representation has been proven to be of simulator (or the interface) having then to put them in the
wide application and to allow for considerable computation appropriate form. It is also advisable to check that reservoir
time savings. Fully compositional fluid representation, when and surface thermodynamic packages have a close behavior.
used in the reservoir simulation, is usually limited to a small When a BO context allows for an accurate flux modeling
number of components (or pseudo components) to keep in both simulations, a much simpler procedure can consist in
acceptable computation time and costs. Using a more detailed delumping the BO reservoir simulation, then passing the
compositional representation (larger number of components) detailed compositions to the surface simulator (in addition to
does not contribute for improving the simulation accuracy in the BO information). The surface simulator uses its internal
terms of volumetric flux. The draw back of simplified fluid BO thermodynamics to compute the flux, and tracks the
representation (BO, or 'coarsely lumped', to refer to compositions, for example for providing a detailed
compositional simulations using a small number of compositional information at the separator level.
components) is the loose of a detailed compositional A method for delumping the BO reservoir simulation25 can
information necessary to the surface process engineering, be derived from the Leibovici method which uses, for each
particularly when distillation columns are used in the component 'i', the following relationship between its
separation process. A smart solution, of strong interest in the equilibrium constant ki and its characterization parameters ai
context of reservoir/surface coupling, has been proposed by and bi in the Peng-Robinson (PR) EOS:
Leibovici23 for restoring a detailed compositional information
from the results of a coarsely lumped compositional Ln(ki) = C0 + C1 ai + C2 bi .......................................... (3)
simulation. The method as described in the paper is not where the C0, C1, C2 coefficients depend only on the phase
applicable to the delumping of Black-Oil simulations. parameters, pressure, and temperature. Discussion is restricted
However, the extension of this method to the delumping of BO to PR EOS for simplifying the presentation, all other details
simulations seems possible (as discussed here after), and being given in the paper 23 and in its references.
should provide a powerful alternative to the use of correlation Figure 7 illustrates the evolution of these coefficients
between gas-oil ratio and produced fluid composition, or to the during the differential vaporization of a reservoir oil modeled
method proposed by Rowney24. The latter, requiring a lot of using PR equation of state.
flash operations using the equation of state (EOS) of the
detailed fluid representation, should be too time consuming to Fig. 7 – Behavior of C0, C1, C2 coefficients
C0
be applicable in the present context. 10
A flexible coupling in terms of compositional information
should allow for using different fluid representations in 8
surface and reservoir models. To establish the continuity 6
between the fluid representations [R] and [S], respectively
4
used in the reservoir and surface simulations, delumping (D)
and lumping (L) operations can be chained, for example using 2
the sequences: 0
(D) (L) (D) (L) 0 50 100 150 200 250
[R] [F] [S] ; [S] [F] [R]
➨ ➨ ➨ ➨ Pressure (bar)
10 C.C. BARROUX, P. DUCHET-SUCHAUX, P. SAMIER, R. NABIL SPE 65159

C1 & C2 4. Most of the present proposals for improving the coupled


0.006 solutions require a greater flexibility in reservoir/surface
data exchange, and some software developments. A key
0.004
C2 point is the collaboration of the software industry. We
0.002
hope that our contribution will help this industry to meet
the petroleum industry expectations in the next
0 C1 developments of their products.
Nomenclature
-0.002 a, b : EOS parameter
0 50 100 150 200 250 A, B : IPR coefficients
Pressure (bar) B : phase formation volume factor
In a Black-Oil context, saturated phase thermodynamic c : component molar fraction in p phase
properties, at a given temperature, are entered in the reservoir C : coefficient
simulation file as a function of saturation pressure either using Cp : specific heat at constant pressure
an analytical correlation or a multi entry table. The behavior of BHP : Bottom Hole Pressure at well reference depth
the C coefficients as shown in Figure 7 is smooth enough to g : gravity acceleration
be entered as supplementary columns of the BO multi entry PI : Productivity Index
table (or to be represented by analytical functions). What is IPR : Influx Performance Relationship
then needed for the delumping of a BO simulation is, in a M : phase mobility
preprocessing step with an external PVT package, to match the N : number (of nodes, of moles, ..)
fluid thermodynamic data (used to build the BO conventional Pb : Grid-block pressure
entries) using an equation of state and a detailed fluid PB : Grid-block equivalent pressure
composition. During this preprocessing, the C coefficients PDOR : Choke or regulator downstream pressure
have to be generated (using the procedure detailed by Pf : Well flowing pressure at perforation depth
Leibovici) at each step of the thermodynamic path used to PLIM : Well limiting pressure in the reservoir model
enter the conventional BO data entries. Then, at any time step Q, Q : volumetric, molar rate
of the BO simulation in any grid cell, the C coefficients can be QMAX : maximum rate constraint
estimated from cell thermodynamic conditions and the T : temperature
extended BO tables, and used for computing the equilibrium T : well/grid block transmissibility
constants. The procedure can then follow what is described in WHP : Well Head Pressure
the Leibovici paper, the only point needing some more Z, Zref : depth, reference depth
discussion being the conversion of the phase volumetric flux z : component molar fraction in bulk mixture
into phase molar flux. This requires the determination of phase ξ : phase molar density
molar densities, these quantities being not computed in the BO ρ : phase density
simulation. One way is to compute the phase molar density as Superscripts
the phase density (provided by the Black-Oil simulation) i : component index
divided by the phase molecular weight obtained using the n : node
phase compositions issued from the LSK delumping o, g, w : oil, gas, water component
algorithm26. st : standard
Conclusions Subscripts
1. A detailed analysis of the problem of coupling two b : loop
independent industrial simulators, driven by the survey of c : component
various actual cases for evidencing the industrial needs, is d, u : downstream, upstream link
presented. Several coupling methods are discussed, and e, s, j : source, sink, junction node
the specific requirements for both reservoir and surface k : perforation index
software are described. l : link
2. A configuration using an overlapping domain between the n : node
surface and the subsurface models extended from the IPR p : phase
to the well head appears to be the most convenient for a o, g, w : oil, gas, water phase
use in both, history match and prediction, contexts of sep : separation
studies.
Acknowledgements
3. Reservoir/surface coupling in a compositional context is
The authors thank TotalFinaElf, Gaz de France and Institut
also discussed. Methods of key interest are those enabling
Francais du Petrole for permission to publish this paper. R.
the delumping of the reservoir simulation results ; a
Brousse, G. Coffe, M.C. Duzan, P. Lemetayer, and O. Ricois
procedure for delumping the results of a Black-Oil
are gratefully acknowledged for fruitful discussions.
reservoir simulation is proposed.
SPE 65159 LINKING RESERVOIR AND SURFACE SIMULATORS: HOW TO IMPROVE THE COUPLED SOLUTIONS 11

References Models", SPE 49001, SPE Annual Technical Conf. and Exh.,
1. Dempsey, J.R., Patterson J.K., Coats K.H., Brill J.P.: "An New Orleans, Louisiana, Sept. 1998.
Efficient Model for Evaluating Gas Field Gathering System 20. Puchyr, P.J.: "A General Gas Reservoir and Gathering System
Design", SPE 3161, JPT (Sept. 1971) 1067-1073. Simulator", JCPT, 30, N° 3 (May-June 1991) 53-60.
2. Startzman, R.A., Brummett, W.M., Ranney, J.C., Emanuel, A.S, 21. Litvak, M.L., Macdonald C.J., Darlow B.L.: " Validation and
Toronyi, R.M.: "Computer Combines Offshore Facilities and Automatic Tuning of Integrated Reservoir and Surface Pipeline
Reservoir Forecasts ", Petroleum Engineer (May 1977), 65-74. Network Models ", SPE 56621, SPE Annual Technical Conf. and
3. Emanuel, A.S, Ranney, J.C.: "Studies of Offshore Reservoir Exh., Houston, Texas, Oct. 1999, 297-305.
With an Interfaced Reservoir/Piping Network Simulator", Trans. 22. Hepguler, F. et al.: "Applications of a Field Surface &
AIME, 271 (JPT, Mars 1981) 399-406. Production Network Simulator with a Reservoir Simulator", SPE
4. Breaux, E.J. et al.: "Application of Reservoir Simulator 38007, Reservoir Simulation Symp., Dallas, June 1997, 285-286.
Interfaced with a Surface Facility Network: A Case History", 23. Leibovici, C. F., Barker, J. W., Wache D. : " Method for
Trans. AIME, 279 (SPEJ Jun. 1985) 397-404. Delumping the Results of Compositional Reservoir Simulation ",
5. Mohamed, D.A., Steffensen R.J.: "An Efficient Reservoir- SPE 64001, SPEJ 5 (2), June 2000, p. 227 (from SPE 49068).
Coupled Gas-Gathering System Simulator", SPE 8333, SPE 54th 24. Rowney, J.D., Clonts M.D.: US Patent N° 5710726.
Annual Fall Conf. & Exh., Dallas, Sept. 1979. 25. Patent pending.
6. Stoisits, R.F., Batesole, E.C. et al.: "Application of Nonlinear 26. Leibovici, C. F., Stenby, E.H., Knudsen, K.: "A Consistent
Adaptive Modeling of Rigorous Representation of Production Procedure for Pseudo-Component Delumping", Fluid Phase
Facilities in Reservoir Simulation", SPE 24898, SPE Annual Equilibria, 1996, 117, 225-232.
Technical Conf., Washington, Oct. 1992, 425-434. 27. Osiadacz A.J.: Simulation and Analysis of Gas Networks, Gulf
7. Lyons, S.L. et al.:"Integrated Management of Multiple-Reservoir Publishing Company (1987)
Field Development", JPT, 47, N° 12 (Dec. 1995), 1075.
8. De Moegen, H., Nabil, R., Lehuen, P., Sonier, F.: "Well-Surface Appendix A : Closure of a network problem
Connection Control for Gas Reservoir Simulator", SPE 35625, For setting the problem in a simple way, a single connected
Gas Technical Conf., Calgary, May 1996. network is considered. Its topology is described by nodes and
9. Hooi, H.R., Goobie, L., Cook, R., Choi, J.: "The Integrated Team links. Nodes are either boundary nodes (source, Ne nodes, or
Approach to the Optimization of a Mature Gas Field", SPE
sink, Ns nodes) or junction nodes (Nj nodes). A boundary node
26144, SPE Gas Technical Symp., Calgary, June 1993, 73-80.
10. Tingas, J., Frimpong, R., Liou, J.: " Integrated Reservoir and is connected to a single link, this restriction being adopted in
Surface Network Simulation in Reservoir Management of most of the commercial network simulators. A link is an
Southern North Sea Gas Reservoirs ", SPE 50635, SPE European oriented segment joining two nodes, in which can be
Petroleum Conf., The Hague, Netherlands, Oct. 1998, 51. represented a succession of various devices such as flow lines,
11. Litvak, M.L. and Darlow B.L.: " Surface Network and Well compressors, chokes, ..
Tubinghead Pressure Constraints in Compositional Simulation ", The total number of nodes is:
SPE 29125, 13th SPE Symp. on Reservoir Simulation, San
Antonio, TX, U.S.A., Feb. 1995, 325-336. Nn = Ne + Ns + Nj .................................................... (A-1)
12. Litvak, M.L. et al.: "Integration of Prudhoe Bay Surface Pipeline From the graph theory (see for example ref. 27), it can be
Network and Full Field Reservoir Models", SPE 38885, SPE
demonstrated that the number of links is given by:
Annual Technical Conf., Oct. 1997, 435-443.
13. Litvak, M.L. and Wang C.H.: " Integrated Reservoir and Surface Nl = Nn − 1 + Nb .................................................... (A-2),
Pipeline Network Compositional Simulations ", SPE 48859, SPE
International Oil & Gas Conf. and Exh., Beijing, China, Nov. Nb being the number of loops in the network.
1998, 297-305. Another useful relationship is:
14. Trick, M.D. et al.: "Gas Field Deliverability Forecasting: A
Nn Nn
Coupled Reservoir Simulator and Surface Facilities Model", CIM
94-62, 45th Annual Conf. of CIM, Calgary, June 1994. å Ndn = å Nun = Nl ............................................ (A-3),
15. Hepguler, F., Barua, S., Bard, W.: "Integration of a Field Surface n=1 n=1
and Production Network with a Reservoir Simulator", SPE Nd and Nun being respectively the number of downstream
n
38937, SPE Computer Applications (June 1997), 88-93. and upstream links connected to the node n.
16. Trick, M.D.: "A Different Approach to Coupling a Reservoir
Simulator with a Surface Facilities Model", SPE 40001, SPE Gas Flow sign in a link may be positive or negative (the link
Technology Symp., Calgary, March 1998. orientation is not necessarily the flow orientation into the
17. Nacul, E.C., Lepetre, C., Pedrosa, O.A., Girard, P., and Aziz, K.: link). By convention, one sets that at a junction node, at least
"Efficient Use of Domain Decomposition and Local Grid one link is oriented towards the node (centripetal link
Refinement in Reservoir Simulation ", SPE 20741, 65th Annual regarding the node position) and one link is centrifuge.
SPE Technical Conf. & Exh., New Orleans, 23-26 Sept. 1990
18. Schiozer, D.J, Aziz, K: "Use of Domain Decomposition for
Simultaneous Simulation of Reservoir and Surface Facilities",
SPE 27876, SPE Western Regional Meeting, Long Beach, March
1994, 269-278.
19. Byer, T. J. ,Edwards, M.l G. ,Aziz, K.: " Preconditioned Newton
Methods for Fully Coupled Reservoir and Surface Facility
12 C.C. BARROUX, P. DUCHET-SUCHAUX, P. SAMIER, R. NABIL SPE 65159

The main unknowns are listed in Table A-1 :


Table A-4: Number of unknowns per type For composition calculations:
Unknown Number - [(Nc − 1) × Nj] equations for component mass
Node pressure Nn conservation at each junction node ;
Node temperature Nn - [(Nc − 1) × Nn] equations for relating the bulk fluid
Fluid molar composition at node(*) Nn × (Nc – 1) composition at node n and fluid composition in the links
arriving in or leaving the node n
Temperature at link beginning Nl
- [(Nc − 1) × (Ndn − 1)] equations for constraining the
Temperature at link end Nl
composition in the downstream links of the junction node
Link global molar rate Nl n ; summing on all junction nodes give a number of
Link global molar composition Nl × (Nc – 1) [(Nc − 1) × (Nl – Nj – Ne)] equations.
(*): Nc, number of components. Closure relationship: For pressure/rate calculations:
åz = 1 , is used.
i
- Nl pressure drop equations allowing for calculating
i pressure gradient along the links ;
As the network is steady-state, there is no possibility of - Nj equations for global mass conservation at each junction
fluid accumulation in the nodes. node n ;
At any standard junction node (no operation of fluid - Per separator node n with Ndn downstream branches, and
separation at nodes), the fluid composition is the same in each given that there is no accumulation possible in any node,
downstream link: it results from the component mixing of one has to consider Nsepn supplementary equations
fluids arriving from upstream links, mixing being constrained constraining the downstream global molar rates, Nsepn
by mass conservation equations. being equal to (Ndn − 1). For example, a flash operation
In an analogous way, per enthalpy conservation and for gas/oil separation, fixing the molar liquid and vapor
assuming that no external energy flux is arriving at node n, the fractions, constrains the liquid/vapor rate ratio at each
temperature is the same at the upstream extremity of each separator exit. Summing on the number of separator nodes
downstream link: the temperature value is the node gives the total number of elementary operations of
temperature ( See Figure A-1). separation Nsep.
The system of equations is composed of:
For temperature calculations: The following table recapitulates, per type of unknown, the
- Nl energy balance equations allowing for calculating number of unknowns and the number of available equations:
temperature gradient along the links ;
- Nj equations (a1) for enthalpy conservation at each Table A-5: Number of unknowns and equations
junction node n ;
Unknown Nbr of Unknowns Nbr of Equations
- Nn equations (a2) for deriving node temperatures ;
Temp. 2 × N l + Nn 2 × N l + Nn – N e
- (Ndn − 1) equations (a3) for constraining the upstream
temperature of downstream links per junction node ; Compo. (Nc − 1) × (Nl + Nn) (Nc − 1) × (Nl + Nn – Ne)
summing on all junction nodes give a number of Pres./Rate Nn + Nl Nn + Nl + Nsep − Ne − Ns
(Nl – Nj – Ne) equations.
The difference corresponds to the number of conditions
Fig. A-1 – Enthalpy equations at node (n) necessary for the problem closure:

l3 Link orientation Table A-6: Closure conditions


Flow orientation Unknown Number of conditions
Temp. Ne
l5
Node n T° Tn Compo. (Nc − 1) × Ne
n Pres./Rate Ne + Ns − Nsep
Link T° at extremity n
l2 T1 n l1 T1n
l4 It can be noted that the number of loops does not appear in
l2 T2n
l1 the final enumeration.
l3 T3n
As many temperature conditions as the number of source
l4 T4n nodes is requested for the closure of the network problem: it is


$%&

Σ Ql Cpl Tln = 0 (a1) l5 T5n why it is chosen in the network simulators to request for a
! 

l∈n
temperature boundary condition at each source node. In the

$%&

$%&

Σ |Ql | Cpl Tln


l∈n
− Σ |Ql | Cpl
l∈n
Tn = 0 (a2)
(a1)
(a2) Þ(T4n = Tn)
same way, a composition has to be provided by the user at
each source node.
!
!

(a3) To represent one 'two downstream branches' separator in a


T3n = Tn (a3) network problem, neither a pressure nor a rate has to be fixed
SPE 65159 LINKING RESERVOIR AND SURFACE SIMULATORS: HOW TO IMPROVE THE COUPLED SOLUTIONS 13

at one of the boundary nodes. If, as usual in the commercially å Tk æå Mpk ξpk cpikö (Pbk − Pfk)
available network simulators, a pressure (or rate) boundary j∈w
èp ø
condition has to be fixed at each boundary node, then there is zi = ..................... (B-7)
no possibility to represent a separator as a node with two å Tk æå Mpk ξpkö (Pbk − Pfk)
downstream branches. This is why the separator is represented
j∈w
èp ø
as a link device, not as a node device, in such simulators. Note that, for applying these equations in a strict BO
It is also to be noted that this constraint does not exist for a context, one has just to define the phase molar densities from:
transient network problem since accumulation in a node, such
ξsto + Rsk ξstg ξstg ξwst
as a separator node, can be represented. ξok = ; ξgk = ; ξwk = ... (B-8)
Another point is the problem of a change in flow direction Bok Bgk Bwk
at a boundary point declared as a source point, this source and the component molar fractions in each phase as:
point becoming a sink point. At the time of status switch, the
network problem is over constrained, since the number Ne of ξsto Rsk ξsto
cook = g
st ; cok = ; cg = 1 ; cwwk = 1
source points has lost one unit ; it is then necessary to cancel ξsto + Rsk ξ g ξ o + Rsk ξstg gk
st

the temperature and composition conditions at the boundary (B-9)


node where the flow direction is reversed. Such a process is The standard way for calculating the quantity (Pfk − BHP),
necessary to allow for handling the cross-flow in the wells. i.e. the pressure drop between the well datum depth and the
The enumeration above holds for a given state of the reference depth of the kth active perforation, is to consider only
network, i.e. at fixed choke openings, at fixed rotation speeds the hydrostatic pressure gradients (the dynamic pressure drops
for compressors, pumps, .. The number of degrees of freedom are then neglected) ; these are computed from a sum of
corresponds to the number of floating regulations in choke (ρhk g ∆Zk) terms in which the bulk phase densities ρhk in the
openings, in compressor or pump rotation speeds. The network hole are based on the bottom hole pressure BHP corresponding
problem will have a single solution if one adds a number of to the well working point calculated at the beginning of the
new constraints equal to the number of degrees of freedom. If time step.
the number of additional constraints is smaller, the problem The thermodynamic conditions, in particular the pressure,
has multiple solutions, the selection of a particular one being used to compute the phase molar densities are those of the grid
either deterministic, or from an optimization process. block ‘k’ if the 'k' perforation is producing, and those
prevailing in the well at the perforation depth in the opposite
Appendix B : Generalization of IPR case. It follows that, in a coupled injection problem, as IPR
Using the standard well/reservoir connection, the molar rate of coefficients mix variables from both simulations, IPR
phase ‘p’ between the grid block ‘k’ (corresponding to the kth overlapping may be of delicate use.
active perforation of the well w) and the well bore is given by: As long as the bulk phase densities are handled explicitly
Qpk = Tk Mpk ξpk (Pbk − Pfk) ......................................(B-1) (no update of (ρhk g ∆Zk) terms during the Newton iterations
of the reservoir simulator, even though BHP values are
Summing on all phases, the molar rate of component 'i' is
revised), and as long as only producing perforations are
given by:
considered, A and B do not depend directly on BHP inside a
Qik = Tk æå Mpk ξpk cpikö (Pbk − Pfk) ........................(B-2) given time step.
èp ø A way for establishing a link with Hepguler method is to
introduce two IPR coefficients per component 'i' as:
Summing on all active perforations, the molar rate of
component 'i' at the well level is given by: Ai = A zi ; Bi = B zi ................................................. (B-10)
Reminding that, in a BO context, phase (oil, water, gas)
Qi = å Tk æå Mpk ξpk cpikö (Pbk − Pfk) ....................(B-3) volumetric rates at standard conditions are equivalent to
k∈w
èp ø component (oil, water, gas) molar rates, one can establish the
It can be rewritten as: Qi = Q zi , Q being given by: correspondence between the 'phase' IPR coefficients Ap and Bp
used by Hepguler and the 'component' IPR coefficients
Q = A − B BHP .........................................................(B-4)
introduced here above. The ratio Ai/Bi (≡ A/B), the same for
with: all the components, is appropriate for use in the network
A = å Tk æå Mpk ξpkö (Pbk − Pfk + BHP) .....................(B-5) simulator as an equivalent reservoir pressure boundary
j∈w èp ø condition, and the reservoir/surface coupling can be realized at
the time step level.
B = å Tk æå Mpk ξpkö ....................................................(B-6) When sophisticated options for wellbore calculations (as
j∈w èp ø implicit fluid densities) are activated in the reservoir model,
and more particularly when several perforated grid blocks
and: contribute to the well production, the coupling and update of
the IPR coefficients have to be realized at the Newton iteration
14 C.C. BARROUX, P. DUCHET-SUCHAUX, P. SAMIER, R. NABIL SPE 65159

level, given the dependence of A and B on BHP inside the


time step.
Another method for generating the IPR coefficients is to
use the first order derivatives of component molar rates versus
well bottom hole pressure, calculated in the reservoir simulator
when the wells are handled implicitly in the simulation:

∂Qi ∂Qi
Ai = Qi0 − BHP0 ; Bi = − ........... (B-11)
∂BHP 0
∂BHP 0

This 'tangent IPR' method is more general than the first


presented procedure. It is applicable in particular when non
linear inflow performance relationships are used in the
reservoir simulator, noting that the IPR coefficients update
may be necessary at each Newton iteration.
Usually, for well implicitness, the '∂Q/∂BHP' derivatives
are computed for each well working with a rate constraint, and
not computed for wells working with a pressure constraint.
It is thought to be the reason why, in the coupling approach
described by Trick (where the surface simulator passes bottom
hole pressures to the reservoir simulator), it is chosen, in a
procedure for remedying flow rate oscillations, to compute the
IPR coefficients from a numerical derivation at the interface
level of the '∂Q/∂BHP' quantities, with no mention concerning
the possibility of using IPR coefficients passed from ECLIPSE
to FORGAS. A small development in the reservoir simulator
for a call to the subroutine devoted to the computation of the
'∂Q/∂BHP' derivatives, whatever the type of constraint (rate or
pressure) applied to the wells, would be helpful to the Trick
approach.
To constrain the wells in a reservoir simulation (explicit in
bulk phase density into the well bore), it is more
advantageous, in a coupling context where wells have multiple
perforations, to fix BHP instead of well rate, since, in the first
case, the calculation of the flowing pressures in the tubing at
each perforation depth is done only one time at the beginning
of the time step. Once again, the '∂Q/∂BHP' derivatives, for
determining the 'tangent IPR' coefficients to be passed to the
surface simulator, have to be computed although they are not
to be used by the reservoir simulator for this type of boundary
condition.
The Hepguler approach, where rates computed by the
surface simulator are used for constraining the reservoir
simulation, does not require any modification of the reservoir
simulator (the '∂Q/∂BHP' derivatives are available since used
for the reservoir computations). But it appears to be optimal
only when there is a single active perforation per well: in that
simple case, rate data put in a consistent form can be directly
introduced as source terms in the reservoir model equations.

SI Metric Conversion Factors


bar × 1.0* E+05 = Pa
* Conversion Factor is exact

You might also like