Sei sulla pagina 1di 12

SPE-182704-MS

A Novel IPR Calculation Technique to Reduce Oscillations in Time-Lagged


Network-Reservoir Coupled Modeling Using Analytical Scaling and Fast
Marching Method

Yanbin Zhang, Gaurav Seth, and Jianping Chen, Chevron ETC

Copyright 2017, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Reservoir Simulation Conference held in Montgomery, TX, USA, 20–22 February 2017.

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 have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
Surface network and subsurface reservoir coupled modeling adds value to asset development assessments
when gathering networks and processing units impose significant constraints on production and injection
rates. However, it is often difficult and time-consuming to reduce or avoid oscillations in production
forecasts that often occur when the surface network and reservoir model are coupled together time step
lagged. We present a new technique for IPR calculations which significantly reduces or eliminates these
oscillations and provides smoother more realistic production forecasts from time-lagged coupled modeling
without significant user intervention. This technique has three steps. First, the conventional IPR curve is
calculated by solving the well model in the reservoir simulator multiple times over a range of flow rates
or bottom-hole pressures (BHP). Second, a dynamic "region of influence" (ROI) is generated for each well
using the Fast Marching Method (FMM). Third, the average pressure in the ROI is used to analytically scale
the conventional IPR curve. The analytical scaling process honors the well operating point in the reservoir
model at the time of coupling, maintains its existing non-linear shape, and shifts the no-flow BHP of the
conventional IPR curve to the average pressure in the ROI.
The conventional IPR curve mentioned earlier only considers the cells in which the well is completed and
assumes no change in pressure in these cells until the next coupling time step. As a result, it typically predicts
well productivity indices (PI) that are too high so that the coupled model forecasts oscillatory production
rates over time. Previous studies have identified this problem and proposed improvements by considering
regions around wells for IPR calculations. These methods usually require advancing and winding back
flow simulations in reservoir or sub-domain models multiple times between coupling time steps. This
implementation requires access to reservoir simulation code and the run time cost can be high and in some
cases prohibitive. In contrast, the new technique presented in this paper can be implemented independent
from the reservoir simulation code and the run time cost is low. In addition, previous studies provide no
clear guideline on how to define the region for a particular well and how the region should change over
time. In this paper, we use FMM to define the ROI dynamically around wells, so that its extent depends on
simulation time, the coupling time interval and reservoir heterogeneity. We believe the ROI obtained this
way reasonably captures the region of the reservoir which most significantly influences well performance
until the next coupling time step. Therefore, the average pressure in the ROI provides a good estimate of
2 SPE-182704-MS

the no-flow BHP which is used as the basis to adjust the IPR curve. We demonstrate the application of this
new technique in a field example.

Introduction
Surface network and subsurface reservoir coupled modeling adds value to asset development assessments
when gathering networks and processing units impose significant constraints on production and injection
rates. Particularly in the current low oil price environment, it is critical for major deepwater assets to
carefully evaluate or re-evaluate their facility design in order to save cost and maximize economic value.
A fit-for-purpose network/reservoir coupled model often provide vital information to differentiate between
alternative facility designs, identify potential operational problems for early intervention/mitigation, and
develop strategies to maximize recovery.
The coupling between surface network and reservoir simulation models can be achieved at different
levels. At one end of the spectrum, the two models can be fully integrated and the entire system is solved
implicitly for every simulation time steps (Coats, et al. 2004, Fleming and Wong 2013, Cao, et al. 2015)). At
the other end of the spectrum, the two models can be coupled time-step lagged without iteration (Guyaguler
and Ghorayeb 2006, Ghorayeb, et al. 2003). Intermediate level of coupling can be achieved by iteratively
solving the two models at either time step or Newton level. Although fully integrated coupling offers the best
solution in theory, its implementation is quite challenging and currently not widely available in commercial
simulation packages. On the other hand, time-lagged coupled modeling is straight-forward to implement
and offers great flexibility in software selection. Detailed discussion of the advantages and disadvantages
of various coupling schemes can be found in the literature (Guyaguler, Zapata, et al. 2011, Cao, et al.
2015). This paper will only focus on the time-lagged coupled modeling, which often suffers from severe
oscillations.
In a time-lagged coupled model, at any coupling time step, an inflow performance relationship (IPR) is
generated for each well according to the subsurface flow simulation. These IPR curves are used as boundary
conditions to the network model to fully represent the subsurface flow. Once the network model is solved
to find the new operating point for each well, either the flow rate or the bottomhole pressure (BHP) from
the new operating point is imposed on the particular well in the reservoir model to continue subsurface
flow simulation until the next coupling time step. The reservoir model and the network model are solved
sequentially without iteration. The passing of information between the two models is purely through IPR
curves (from reservoir model to network model) and well operating constraints (from network model to
reservoir model). In terms of time stepping sequences, the reservoir model may take multiple simulator
time steps between two subsequent coupling time steps (Figure 1). Traditionally IPR curve is calculated
by solving the "well model" in the subsurface simulation by varying flow rate or BHP. The "well model"
is composed of many well-to-cell connections, and each connection is modeled by the Peaceman Equation
(Peaceman 1977).
SPE-182704-MS 3

Figure 1—The process of time-lagged surface network and subsurface reservoir coupled modeling.

The numerical instability of time-lagged coupled models has been studied in the past by several authors.
It has long been realized that the traditionally calculated IPR at the beginning of the coupling time step may
not be representative of the IPR during the entire coupling time interval. Specifically, the traditional IPR
calculation only considers the well and its completions cells representing an instantaneous behavior, while
more stable IPRs should be based on well influence area over the subsequent time step. Guyaguler et al.
(2011) developed an IPR calculation technique based on sub-domain simulations. Liang and Rubin (2014)
proposed a method by simulating simultaneous flow tests of all reservoir wells. In both of these methods,
additional reservoir simulation runs are needed either on the fly or before the coupled simulation in order to
obtain more stable IPRs. These extra simulations usually add significant computational cost and may require
access to reservoir simulator's internal code to be implemented. In this paper, we proposed a novel stabilized
IPR calculation method based on a simple analytical scaling and the Fast Marching Method (FMM).
FMM has been an active research topic in recently years as a novel technique for drainage volume
calculation for unconventional reservoirs (Datta-Gupta, et al. 2011). It has been shown that the classic
concept of radius of investigation in pressure transient analysis for homogeneous reservoirs can be extended
to heterogeneous reservoirs by using FMM to solve for the pressure front propagation. In this paper, we
apply FMM to define the "region of influence" (ROI) for each well from which the more stabilized IPR
curve is derived.
This paper is organized as follows. First we analyze one of the most common causes of the oscillation
problem for time-lagged coupled modeling by using an idealized example. This analysis reveals the
underlining assumption behind the traditional IPR calculation and demonstrates that in some cases it will
not be suitable for time-lagged coupled modeling. Next, we develop the methodology by introducing a new
assumption more suitable for time-lagged coupled modeling. Then we briefly discuss the implementation
of the new method and show its application in a field example. Conclusions are made in the final section.
4 SPE-182704-MS

Problem Analysis
To understand one of the most common causes of oscillations for time-lagged coupled modeling, let's
consider the flow conditions of two wells, red and blue, as illustrated in Figure 2. The two wells both produce
at the same BHP and flow rate Q in steady-state, single-phase radial flow in their respective homogeneous
reservoirs with the same radius of reservoir boundary re and the same boundary pressure Pe. Both reservoirs
are discretized into grid blocks with a cell size ΔX (re ≫ ΔX). The blue well produces from a higher
permeability reservoir but has a positive skin factor that impairs productivity, while the red well produces
from a lower permeability reservoir but has a negative skin factor that enhances productivity. Using the
steady-state radial flow equation, the productivity indexes (PI) of the two wells should be identical and can
be calculated as

(1)

Figure 2—The flow conditions of two wells, red and blue.

The subscript "ss" stands for "steady state". The blue and red dots in Figure 2 show the pressure profiles
for the two wells overlaid on each other. In the blue well case, most pressure drop occurs near the wellbore
in the completion cell, while in the red well case, most pressure drop occurs deeper in the reservoir and
the pressure drop from the completion cell to the BHP is very small. Consequently, the traditional IPR
curves obtained by solving the well model in the reservoir simulation are quite different for the two wells
as illustrated in Figure 3(a). Specifically, the zero-rate pressures on the IPR curves, which are equal to the
completion cell pressures, Pcell, are different for the two wells, and the slopes of the IPR curves, which are
the negative inverse of the wells' PIs, are also different for the two wells.
SPE-182704-MS 5

Figure 3—IPR curves obtained by the traditional way, i.e., solving well model in the reservoir simulator. Left
plot shows the blue well case; right plot shows the red well case. Black curves on both plots show the same
flow curve generated by the network model. Black dots on both plots show the operating points of both wells.

At coupling time t, let us introduce a small disturbance to the network model by adding a small choke
ΔP. The network flow curve shifts upward. A new operating point can be obtained by the intersection of the
well's IPR curve to the new network flow curve. Figure 3(b) shows the solution of the new operating points
for both wells. The flow rate of the red well is choked back more drastically than the blue well because the
slope of the red well's IPR curve is much smaller. After the coupling is done, new well producing constraints
are imposed on the red and blue wells respectively to continue reservoir simulation until the next coupling
time step t + ΔT.
Because the well producing constraints are changed after the coupling at time t, the reservoirs will respond
to the new boundary conditions (well is the inner boundary of the reservoir), by adjusting the pressure
profiles in the reservoir. For the red well, because the production is severely choked back for the time period
ΔT, if ΔT is relatively large, the completion cell pressure Pcell at the next coupling time step t + ΔT will
have been built up to be much higher than the previous value (and much closer to Pe). Therefore, at the
6 SPE-182704-MS

next coupling time step t + ΔT, the red well's IPR curve will shift up significantly as shown in Figure 3(c).
Similar response of the IPR curve will occur for the blue well, but to a much less magnitude for the same
coupling time interval ΔT.
At the coupling time step t + ΔT, a new operating point is obtained for both the red and the blue wells by
the intersection of the new IPR curves and the new network flow curve. As illustrated by Figure 3(c), the
flow rate of the red well will swing to the other side of the plot and will be much higher than the original
steady-state flow rate Q. The reservoir response of such high flow rate for the red well will again lead back
to a situation illustrated in Figure 3(b) and, thus, causes oscillations in subsequent coupling time steps.
Oscillations occur for the blue well too, but the magnitude is much less and probably will dampen and
diminish after a few coupling time steps.
Remember that the steady-state productivity indexes for the two wells are the same as given by Equation
(1). We cannot help but to ask why the traditional IPR calculation produces two different answers for the two
wells and why for the same coupling time interval ΔT, the red well tends to oscillate more than the blue well.
To answer these questions, it is necessary to scrutinize the assumption behind the traditional IPR
calculation. The well model in the reservoir simulation is essentially based on the Peaceman Equation,
which assumes steady-state radial flow (Peaceman 1977). Thus, the traditional IPR calculation also assumes
steady state with the completion cell pressure Pcell remaining constant from time t to t + ΔT. This assumption
is only perfectly true in an "instantaneous" sense and works satisfactorily only if Pcell does not change too
much between time t to t + ΔT no matter what the new operating constraint is set after the coupling is done.
For the same ΔT, the red well case is more susceptible to violate this assumption than the blue well case.
This is because most pressure drop is outside the completion cell due to low permeability in the reservoir
and negative skin factor of the red well. Another observation is that the productivity index associated with
the traditional IPR can be calculated as

(2)

The subscript "in" stands for "instantaneous". The "instantaneous" PI of the red well is much larger than
the blue well and it is also much larger than the steady-state PI in Equation (1). This "instantaneous" PI is
only applicable in a short-term transient period. For example, in Figure 2 for the red well, if BHP suddenly
increases slightly above Pcell at time t, the well immediately cannot flow. However, once the well is shut-in,
the pressure near the well will build up, and the well will gradually start to flow again and approach a new
steady state with a flow rate slightly less than the previous steady-state flow rate Q. If we are not interested
in the short transient flow period and only care about long-term production, the "true" steady-state IPR
curve should be used for the coupling as shown in Figure 4. The underlining assumption is still steady state,
but different from the traditional IPR calculation, it is the reservoir boundary pressure Pe rather than the
completion cell pressure Pcell that remains constant. Using the "true" steady-state IPR curve, the coupling
solution for both the red and the blue wells becomes identical and unconditionally stable even for ΔT → ∞.
However, any short-term transient behavior of the wells cannot be captured under such an assumption.
SPE-182704-MS 7

Figure 4—Steady-state IPR curve can be used for both blue and red wells in the example, which
gives a one-time stable solution without any oscillation even if coupling time interval ΔT →
∞. However, any transient behavior for the wells cannot be captured by using this IPR curve.

By carefully examining the assumptions behind IPR calculations for the simple steady state example, it
is clear that the traditional IPR curves in Figure 3(a) and the true steady state IPR curve in Figure 4 are
applicable for different situations. To model short-term transient behavior, the traditional IPR curve should
be used; to predict long-term production, the true steady-state IPR curve should be used. For a time-lagged
coupling problem with a given coupling time interval ΔT, we need a "fit-for-purpose" IPR curve in order
to produce smooth and physical production forecasts. In the following section, we will develop this "fit-
for-purpose" IPR curve.

Methodology
The subsequent development is based on the following simple assumption: At coupling time t, each well is
treated as if it is in a "steady stat"e until the next coupling time t + ΔT. The rationale behind this assumption
can be shown as follows:

• The limitation of time-lagged coupled modeling dictates that any short-term transient behavior
below the time resolution ΔT cannot be resolved. If any short-term transient behavior with a
characteristic time below ΔT is intended to be modeled, the coupling time interval ΔT should be
reduced. (The grid cell size may also need to be reduced accordingly to accurately capture the
transient behavior.)
• ΔT should be small enough so that pressure transient interference between wells in the subsurface
simulation can be ignored during this time period.
• Each well drains its own "reservoir" in the time interval ΔT. However, the change of the average
"reservoir" pressure during ΔT is neglected, i.e., a pseudosteady-state is approximated as a "steady
state". This assumption is, in the worst case scenario, of the same order of approximation as the
time-lag assumption.
• The well model in reservoir simulation which is based on the Peaceman Equation already assumes
steady-state flow between reservoir simulation time steps (from t to t + Δt). Our assumption here
is only an extension of that assumption to the coupling time steps (from t to t + ΔT). When ΔT →
Δt, our assumption falls back to the Peaceman Equation assumption.
• If the assumptions discussed above cannot be satisfied, the time-lagged scheme should be used
with caution. More closely coupled schemes should be considered such as "iteratively lagged" or
"fully implicit".
8 SPE-182704-MS

The "steady-state" assumption implies that the IPR curve should pass through the subsurface operating
point (Q, BHP) at time t, i.e., if well BHP is kept constant, the well should deliver the same production
rate Q "forever" (which means from t to t + ΔT). The "steady-state" assumption also implies that the zero-
rate pressure on the IPR curve should be some "boundary pressure" Pe that we need to estimate. In the
"true" steady-state situation as discussed in the previous section, the boundary pressure Pe is equal to the
well BHP at time t → ∞ if the well is shut-in. Similarly, under the "steady state" assumption defined in this
section, Pe is the well BHP at time t + ΔT if the well is shut-in at time t (see Figure 5). Pe can be obtained
most accurately by simulating a build-up test for the time interval from t to t + ΔT. However, this requires
advancing the subsurface simulation for ΔT time and then winding back the simulation for coupling, which
is computationally expensive. Alternatively, Pe can be approximated as the average pressure at time t in the
ROI, which is enclosed by the pressure transient front propagating into the reservoir in the time interval
ΔT. Because all calculation can be done at time t, there is no need to advance and wind back subsurface
simulation.

Figure 5—Estimate the "boundary pressure" Pe under the "steady state" assumption. If the well
is shut-in at time t, the BHP of the well will build up. At time t + ΔT, the BHP value would be Pe.

Once the "boundary pressure" Pe is obtained, we analytically scale the traditional IPR curve according
to the following equation for oil producers (as illustrated in Figure 6)

(3)

Figure 6—Analytical scaling of the traditional IPR curve to obtain the new IPR curve.

This analytical scaling has the following characteristics:


1. Anchor at the current subsurface operating point
2. Keep existing non-linear curve shape
3. Zero-rate pressure on the IPR curve is scaled from Pcell to Pe
SPE-182704-MS 9

4. Fall back to the traditional IPR for small ΔT (when Pe approaches Pcell)
For water injectors, the analytical scaling equation is similar. For gas wells, "pressure squared" should
be used to substitute pressure terms in the analytical scaling calculation.
As mentioned earlier, the ROI used to obtain Pe is defined in terms of the pressure transient front
propagation. In homogeneous radial flow, this is simply the "radius of investigation" concept widely used
in pressure transient analysis (Lee 1982). Extension of this concept to heterogeneous reservoirs leads to the
Eikonal Equation (Vasco 2000)
(4)
where

(5)

In Equation (4), τ is called the diffusive time-of-flight, α is the diffusivity which depends on permeability
k, porosity ϕ, viscosity μ, and total compressibility ct. The Fast Marching Method (FMM) is an efficient
algorithm to solve the Eikonal Equation (Sethian 1996). Once Equation (4) is solved with FMM to obtain
the diffusive time-of-flight τ for each cell (Zhang, et al. 2013), the ROI used to obtain Pe can be defined as
(6)
where
(7)
Equation (7) is an approximation based on radial flow assumption (Zhang et al. 2013). To summarize,
the new IPR calculation method has three steps:
1. Obtain the traditional IPR curve
2. Generate ROI using FMM and obtain Pe as the average pressure in the ROI
3. Apply Equation (3) to scale the IPR curve according to Pe

Implementation and Application


The new IPR calculation method has been implemented in the Chevron in-house coupled modeling
controller. The controller is a piece of software facilitating the communication between a reservoir simulator
and a surface network simulator for coupled modeling. The controller calls a FMM subroutine developed
using the Python scripting language interface provided by the reservoir simulator. The FMM subroutine
will define/update the ROI for each well in the simulator. Then the controller obtains the traditional IPR
curves and the average pressure of ROIs from the reservoir simulator for the analytical scaling calculation.
A field example using the new IPR calculation method follows.
The reservoir model has relatively low permeability (~15 md) and highly under-saturated oil. Nine oil
producers are to be drilled sequentially. All wells are hydraulically fractured with large negative skin factors.
The network model contains two flow lines with pumps starting at different times. A time-lagged coupled
model was built and simulated using the Chevron in-house controller with a commercial reservoir simulator
and a surface network simulator from different vendors. Coupling time interval is one month. Figure 7
shows the oil production forecasts for four out of the nine producers (P02, P03, P04, and P05) using the
traditional IPR calculation. The results are unphysical due to large oscillations. Previous efforts to reduce
oscillations were to develop heuristics when solving the network model, e.g. constraining well flow rate
change by no more than 5% from the previous coupling time step. With the new IPR calculation method,
10 SPE-182704-MS

no heuristics is needed to obtain smooth more realistic oil production forecasts as shown in Figure 8. It also
better captures the response from the pump startup when production rates suddenly increase for all wells
around 2,250 days.

Figure 7—Oil production forecasts for wells P02, P03, P04, and P05 using the traditional IPR curve.

Figure 8—Oil production forecasts for wells P02, P03, P04, and P05 using the traditional IPR curve with heuristics
to reduce oscillation (dashed lines) compared with using the new IPR curve without any heuristics (solid lines).

Figure 9 shows an example of the traditional IPR curve compared with the new IPR curve. The traditional
IPR curve is almost flat meaning that the associated PI for this well is huge. As discussed earlier, this PI
is only "instantaneous", which may not correspond to the PI that production engineers use in the field.
In contrast, the new IPR curve shows much steeper slope, which gives a PI that is more reasonable for
production engineers to compare with field well performance based on monthly production.
SPE-182704-MS 11

Figure 9—An example of the traditional IPR curve and the new IPR curve extracted from the simulation of the tested model.

Figure 10 shows the ROI generated by the FMM during the coupled simulation. The irregular shape of
the ROI is governed by reservoir heterogeneity. In addition, the ROI also changes dynamically according
to the changing well pattern such as when new wells are drilled. The ROI is generated on the fly and can
be visualized after the simulation is finished.

Figure 10—Dynamic ROI generated using FMM for the field example.
12 SPE-182704-MS

Conclusions
A new IPR calculation method is proposed for time-lagged, network-reservoir coupled modeling to reduce
oscillations in production forecasts. Its basic assumption is: at coupling time t, each well is treated as if it
is in a "steady state" until next coupling time t + ΔT. The new method has the following characteristics:
1. Anchors at the current subsurface operating point
2. Keeps existing non-linear shape of the traditional IPR curve
3. Zero-flow BHP is scaled to the average pressure of ROI dynamically determined by FMM
4. ROI depends on time, coupling time interval, and reservoir properties (heterogeneity)
5. Falls back to the traditional IPR for small ΔT
This new functionality has been implemented in Chevron's in-house software. Application of the new
method has been demonstrated using a field example.

References
Cao, H., P. Samier, H. M. Kalunga, E. Detige, and E. Obi. "A Fully Coupled Network Model, Practical Issues and
Comprehensive Comparison with Other Integrated Models on Field Cases." SPE Reservoir Simulation Symposium.
Houston, Texas: SPE, 2015.
Coats, B. K., G. C. Fleming, J. W. Watts, M. Rame, and G. S. Shiralkar. "A Generalized Wellbore and Surface Facility
Model, Fully Coupled to a Reservoir Simulator." SPE Reservoir Evaluation & Engineering (SPE) 7, no. 02 (2004):
132–142.
Datta-Gupta, A., J. Xie, N. Gupta, M. King, and J. Lee. "Radius of Investigation and its Generalization to Unconventional
Reservoirs." Journal of Petroleum Technology (SPE), July 2011:
Fleming, G., and T. Wong. "Fully Coupled Simulation of Multiple Compositional Reservoirs With a Shared Facility
Network." SPE Reservoir Simulation Symposium. The Woodlands, Texas, USA: SPE, 2013.
Ghorayeb, K., J. Holmes, R. Torrens, and B. Grewal. "A General Purpose Controller for Coupling Multiple Reservoir
Simulations and Surface Facility Networks." SPE Reservoir Simulation Symposium. Houston, Texas: SPE, 2003.
Guyaguler, B., and K. Ghorayeb. "Integrated Optimization of Field Development, Planning, and Operation." SPE Annual
Technical Conference and Exhibition. San Antonio, Texas, USA: SPE, 2006.
Guyaguler, B., V. J. Zapata, H. Cao, H. F. Stamati, and J. A. Holmes. "Near-Well-Subdomain Simulations for Accurate
Inflow-Performance-Relationship Calculation To Improve Stability of Reservoir/Network Coupling." SPE Reservoir
Evaluation & Engineering 14, no. 05 (2011).
Lee, W. J. Well Testing. Richardson, Texas: Seciety of Petroleum Engineers, 1982.
Liang, J., and B. Rubin. "A Semi-Implicit Approach for Integrated Reservoir and Surface-Network Simulation." SPE
Reservoir Evaluation & Engineering 17, no. 04 (2014).
Peaceman, D.W. "Interpretation of Well-Block Pressures in Numerical Reservoir Simulation." SPE Journal 18, no. 3
(1977): 183.
Sethian, J. A. "A Fast Marching Level Set Method for Monotonically Advancing Fronts." Proc. Natl. Acad. Sci. 93, no.
4 (1996): 1591–1595.
Vasco, D. W., Keers, H. and Karasaki, K. "Estimation of Reservoir Properties Using Transient Pressure Data: An
Asymptotic Approach." Water Resour. Res. 36, no. 12 (2000): 3447–3465.
Zhang, Y., C. Yang, M. J. King, and A. Datta-Gupta. "Fast-Marching Methods for Complex Grids and Anisotropic
Permeabilities: Application to Unconventional Reservoirs." SPE Reservoir Simulation Symposium. The Woodlands,
Texas: SPE, 2013.

Potrebbero piacerti anche