Streamline Simulation:
A Technology Update
Akhil Datta-Gupta, SPE, Texas A&M U.
Introduction
Streamline and streamtube methods have been used by the
oil industry for several decades. In recent years, there has
been an increased interest in the technology. This is primarily driven by two factors. First, with the developments
in reservoir characterization, we can now routinely generate high-resolution reservoir models consisting of multimillion cells. This has resulted in a gap between geologic
modeling and flow simulation. Second, with increased
model resolution, there is an increased acknowledgment of
uncertainty. We are interested in assessment of uncertainty
in reservoir description and performance prediction with
multiple plausible reservoir models. Conventional
numerical simulators often are inadequate to satisfy these
needs in a timely fashion.
The primary advantages of streamline methods are
faster computation, improved accuracy (subgrid resolution and reduced numerical dispersion and grid-orientation effects), ability to screen highly detailed geologic
models, quantitative flow visualization, and rapid history
matching or production-data integration into high-resolution reservoir models. The speed and versatility of the
method have led to many novel applications. The disadvantages of streamline models are the difficulties in incorporating complex physical processes and cross-streamline
mechanisms. Streamline models are not a substitute for
conventional grid-based simulators but can play an
important role in bridging the gap between geologic modeling and flow simulation.
Background
Todays streamline simulation was preceded by at least
four other methods for modeling convection-dominated
flow in the reservoir. Line-source/sink methods have
been widely used by the petroleum industry.1,2 These
methods use analytic solutions to the pressure and velocity distribution in the reservoir. The primary limitation of
these methods is the requirement for homogeneous properties and constant reservoir thickness. Streamtube methods are more general and have been applied successfully
for field-scale modeling of waterflooding and miscible
flooding.3-5 In these methods, the flow domain is divided
into a number of streamtubes and fluid-saturation calculations are performed along these streamtubes. However,
the need to keep track of the streamtube geometries can
Copyright 2000 Society of Petroleum Engineers
This is paper SPE 65604. Distinguished Author Series articles are general, descriptive
representations that summarize the state of the art in an area of technology by describing
recent developments for readers who are not specialists in the topics discussed. Written by
individuals recognized as experts in the area, these articles provide key references to more
definitive work and present specific details only to illustrate the technology. Purpose: to
inform the general readership of recent advances in various areas of petroleum engineering.
68
DECEMBER 2000
(a)
(b)
(c)
(d)
(e)
(f)
(g)
Fig. 1Stepwise illustration of streamline simulation:
(a) permeability field, (b) streamlines, (c) time of flight,
(d) water saturation at 0.35 pore volumes injected (PVI),
(e) streamline updating after infill drilling, (f) revised
time of flight, and (g) water saturation at 0.45 PVI.
DECEMBER 2000
(a)
(b)
70
DECEMBER 2000
Fig. 5Comparison of streamline vs. numerical simulation for waterflooding: water saturation at 2.5 PVI and
recovery history.
71
DECEMBER 2000
nectivity at a given time. Swept volume is expected to correlate with recovery regardless of the displacement
process. In the past, permeability connectivity has often
been used to rank geostatistical models. Permeability connectivity may not be very effective because it accounts
only for heterogeneity. In contrast, time-of-flight connectivity accounts for interaction between heterogeneity and
the imposed flow field. Once the models are ranked, we
can choose appropriate quantiles for detailed analysis,
assessment of uncertainty in reservoir description, and
performance predictions (Fig. 7).
Upgridding and Upscaling. The streamline approach is
suitable for upgridding detailed 3D reservoir descriptions.
By upgridding, we mean selecting a coarse simulation grid
based on an existing fine grid for subsequent upscaling calculations. Streamline-based nonuniform coarsening of
fine-scale descriptions naturally places higher resolution
in regions of fast flow and coarser cells throughout the
bulk of the model. Further, the streamline time of flight
provides a quantitative form of flow visualization that can
have a variety of applications in reservoir management and
reservoir characterization. One such application is validation of different upscaling techniques (Fig. 8).
(a)
History-Matching/Production-Data Integration. Streamline methods provide a unique advantage for rapid history-matching or production-data integration into high-resolution reservoir models. Integration of production data
into reservoir models typically requires the solution of an
inverse problem. Streamline models have two advantages.
First, the streamline simulator can serve as an efficient
forward model for the inverse problem. Second, streamline models allow rapid computation of the sensitivity of
the production response with respect to reservoir parameters, such as permeability and porosity. Such sensitivities
quantify change in production response because of a small
change in reservoir properties and form an integral part of
automatic history-matching algorithms. Furthermore, we
can exploit an analogy between streamlines and seismic
ray tracing to pose the history-matching problem similar
to geophysical inversion and use efficient techniques from
(b)
Fig. 9Streamline-based production-data integration into reservoir models: matching of (a) breakthrough response
and (b) corresponding permeability field.
72
DECEMBER 2000
geophysical inverse theory. With streamline methods, production data integration can be carried out in two steps:
(1) matching first arrivals or breakthrough response at
the wells and (2) matching of amplitudesthat is, the production response at all times. Fig. 9 illustrates arrival-time
matching to reconcile geostatistical models to field production history.
Primary Recovery and Compressible Flow. Streamline
methods can be used to define drainage areas or drainage
volumes associated with wells during primary depletion.
Such drainage volumes can be conveniently defined in
terms of a compressible or diffusive time of flight. A
compressible time of flight is based on the observation that
transient-pressure response propagates at a frontal velocity given by the square root of diffusivity. Fig. 10 shows
the drainage area at various times associated with a single
well producing from a heterogeneous medium. For comparison purposes, the figure also shows the commonly
used tracer time of flight for this case. As expected, pressure fronts propagate several orders of magnitude faster
than the tracer fronts.
Solvent Flooding and Compositional Simulation.
Streamline-based techniques have been used to model solvent flooding and various unstable displacements at the
field scale. Most of these techniques rely on analytic or
numerical solution of multicomponent displacement
processes along streamlines. Excellent agreement with
conventional numerical simulators has been demonstrated
with significant speedup in computation. Streamlinebased fully compositional simulation is receiving increasing interest; however, field-scale application has been limited to date.
Discussion
The efficiency of the current generation of streamline simulators is based on their 1D, large-timestep, implicit-pressure/explicit-saturation (IMPES) formulations. In contrast,
DECEMBER 2000
74
DECEMBER 2000
71204.qxd
4/15/08
14:27
Page 26
lines are required for proper modeling, the rank preservation of sweptpore-volume calculation from the first
set of streamlines necessarily becomes
a poor approximation of the recovery
behavior of the system. This point is
demonstrated by progressively allowing more physics (gravity and mobility) to affect the recovery at breakthrough for even a system as simple as
a 3D quarter-five spot as shown in
Fig. 1. The ranking of swept volume
correlates very well with breakthrough recoveries determined by use
of fixed streamlines (Fig. 1a), but progressively degrades as mobility and
gravity affect the recovery responses.
Fig. 1b is a ranking of recovery for a
waterflood, and Fig. 1c is a ranking
for a first-contact miscible (FCM)
flood. Including additional effects,
such as well rate changes and the
desire to predict ultimate recovery,
will necessarily invalidate any ranking
based on swept volume by use of TOF.
Unfortunately in screenings of real
field models, it is almost always necessary to include mobility, gravity, and
changing well conditions to capture
the true range of recovery variability.
History-Matching/ProductionData Integration
Streamline methods do provide a
unique approach to the historymatching problem. As noted, streamlines allow for fast forward simulations of each new flow model, thereby
allowing a larger parameter space to
be investigated or shorter turnaround
times. But the use of sensitivity coefficients as an integral part of automatic
history matching has been shown
only in the context of fixed streamline
problems with no gravity (linear
systems).6,7 Not mentioned is the
more promising assisted history
matching (AHM) method based on
streamlines,8 which has been used
with great success in many recently
published field cases.9-11 The method
exploits the ability of streamlines to
26
MAY 2001
4/15/08
71204.qxd
14:27
Page 27
30
1
1
15
30 1
15
30 1
Recovery at Breakthrough, %
15
27
MAY 2001