SPE 68116
EMpower : New Reservoir Simulation System
B.L. Beckner, SPE, ExxonMobil Upstream Research Co.; J.M. Hutfilz, SPE, ExxonMobil Upstream Research Co.; M.B.
Ray, ExxonMobil Upstream Research Co.; and J.F. Tomich, SPE, ExxonMobil Upstream Research Co.
Copyright 2001, Society of Petroleum Engineers Inc.
This paper was prepared for presentation at the 2001 SPE Middle East Oil Show held in
Bahrain, 17–20 March 2001.
This paper was selected for presentation by an SPE Program Committee following review of
information contained in an abstract submitted by the author(s). Contents of the paper, as
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
for commercial purposes without the written consent of the Society of Petroleum Engineers is
prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300
words; illustrations may not be copied. The abstract must contain conspicuous
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.
Abstract
During the late 1990's, Exxon and Mobil had each
independently developed next-generation reservoir simulation
systems. Both next-generation systems embodied a
substantial number of step-out simulation technologies, which
were extremely complementary. ExxonMobil moved
aggressively to combine the best of both companies'
technologies into one industry-leading simulation system
called EMpower TM. This new simulation system is now being
used to actively manage ExxonMobil’s global resource base.
Key features of this industry-leading simulator are described
in this paper.
The new simulator employs unstructured grids to more
accurately model complex geologic features, near-wellbore
flow, and aquifer support. Algorithms for optimal layering
and flow-based scale-up on unstructured grids are tightly
integrated in the EMpower system. The computations are
performed within the unstructured grid fabric. Interactive
simulation ties together the geologic and reservoir simulation
models with production data yielding high-confidence
forecasts of future performance.
Emphasis is on minimizing the overall turnaround time
between formulation of the simulation problem and generation
of results. A comprehensive graphical user interface provides
reservoir engineers and geoscientists of all skill levels with
easy access to reservoir simulation. The user interface is
designed to facilitate the full range of simulation problems--
from quick screening studies to large, complex field models.
The geoscience and reservoir engineer communicates to the
executing simulation through the user interface, thereby
TM
EMpower is a trademark owned by ExxonMobil Upstream
Research Company.
allowing simulation progress/results to be monitored, paused,
terminated, and/or restarted on command. Results are viewed
within the user interface using spreadsheets, charts, or full 3D
visualization.
The object-oriented design of the new simulator is very
flexible. The reservoir flow model is tightly integrated with
the well and surface facility models for accurate, smoothly
running simulations. Execution of complex well management
strategies is specified in an intuitive, graphical format. For
added efficiency, the simulator also is designed to take
advantage of computing hardware that utilizes multiple
parallel processors.
The capabilities of this next-generation simulation system
are demonstrated through a field study involving complex
geologic features (e.g. non-vertical faults and stratigraphic
pinchouts) and multiple reservoirs connected to a common
production infrastructure.
Introduction
Reservoir simulator development has historically been an
active area of internal research and development at both
Exxon and Mobil.1-6 During the late 1990's, Exxon and Mobil
each independently developed next-generation reservoir
simulation systems. Both next-generation systems employed a
substantial number of step-out simulation technologies, which
were extremely complementary. ExxonMobil moved
aggressively to combine the best of each company’s
technologies into one industry-leading simulation system
called EMpower. This new simulation system is now being used
to actively manage resources within ExxonMobil’s global
asset base. This paper describes the EMpower simulator and
some of its key features.
Potentially the most significant innovation adopted in the
new simulator is unstructured gridding, with computations
performed within the unstructured grid. A number of
researchers have contributed to developments in unstructured
gridding over the past two decades7-9. However, industry
generally has been reluctant to apply this capability to
practical reservoir simulation due in part to concerns about
potential loss in computational efficiency.
Unstructured gridding makes it possible to capture and
honor more geologic and engineering detail in reservoir
simulation models. The past decade has seen significant
improvements in the accuracy and the detail of the data that
are input to reservoir simulators. Examples include detailed
2 B.L. BECKNER, J.M. HUTFILZ, M.B. RAY, J.F. TOMICH
geologic models derived from high quaility 3D seismic,
complex well paths, finer-scale geologic features that clearly
impact reservoir flow, better characterized near-wellbore flow
of complex fluids, and modeling of large-scale interactions
between multiple fields. Expected further improvements in
the quality and quantity of reservoir data over the life of the
new simulator will make unstructured gridding key to accurate
representation of reservoir detail in simulation models.
A key business driver in the design of the simulator is the
need to reduce the overall turnaround time between
formulation of the simulation problem and the generation of
meaningful results. Another key driver is the need to facilitate
a broader range of reservoir simulation applications, including
rapid screening and assessment of large prospects as well as
full-field models for development and depletion planning.
Reservoir engineers and geoscientists of all skill levels are
afforded easy access to this range of applications through a
modern, intuitive Graphical User Interface (GUI). For "quick"
models, the interface facilitates both rapid and efficient data
input as well as sharing of data among users and projects. For
full-field models, design and execution of complex facility
management strategies are specified in an intuitive, graphical
format.
ExxonMobil chose to develop its own proprietary
reservoir simulator primarily to provide a platform for the
rapid implementation of significant technical enhancements
closely aligned with the company's asset base. Experience has
shown that such enhancements continually arise and add
significant value through improved reservoir management
capabilities. The object-oriented design of the new simulator
was selected to provide the necessary flexibility for future
software development.
The simulator tightly integrates flow in the reservoir,
wells, flowlines, and surface facilities to provide a robust
solution for coupled reservoir-facilities flow problems.
Multiple fields with multiple facilities can be simulated using
this integrated flow modeling capability, thus enabling
accurate and efficient solution of problems encompassing
complex reservoir models and complex facility networks. The
practical objectives are to better model the physical constraints
imposed by facility networks and optimize operation and
development of single or multiple fields.
The development of efficient new solvers is required to
allow for the practical solution of simulations employing
unstructured grids and to take advantage of computing
hardware that utilizes multiple parallel processors.
Examples from a current EMpower study are presented to
highlight many of the distinguishing features of the new
simulator. This particular study was chosen from the many
EMpower studies underway at ExxonMobil because it
demonstrates many of the capabilities of this next-generation
simulation system. These include the modeling of complex
geologic features (e.g. sloping faults and pinchouts), the
integration of surface facilities and reservoir models, the use
of flow-based scale averaging for unstructured grids, visual
facility logic, and multiple reservoirs connected to a common
production infrastructure.
SPE 68116
Features and General Specifications
To provide the framework for satisfying ExxonMobil's
reservoir simulation needs for the foreseeable future, the
general architectural specifications for EMpower require that the
simulator be computationally efficient, flexible, extensible,
maintainable and able to be run in parallel on multiple
platforms. These requirements led to a decision to use
objected-oriented programming techniques and the C++
language, which promote the separation of functionality via
non-strict encapsulation. The development platform for
EMpower currently is shared-memory multiprocessor machines,
but the code has been designed for extension to distributed-
memory, multiprocessor architectures. Computational
efficiency concerns about C++ were addressed early in the
project, and those issues were satisfactorily resolved.
EMpower facilitates the modeling of localized phenomena
within a larger model by decomposing the unstructured grid
into computational regions. Using computational regions
allows localized formulation modeling where, for example,
one might employ implicit solution methods near wells and
explicit solution methods away from wells. Unstructured
gridding allows the grid spacing to flexibly vary according to
the modeling need. This localization extends to simultaneous
use of multiple fluid regions such as combining black-oil and
equation-of-state regions.
The unstructured framework also provides an architecture
for directly coupling a surface facility model with the reservoir
model, increasing the computational stability of the entire
system. When viewed in a general unstructured framework,
the integrated facility network is treated as a set of nodes and
connections that is simply a natural extension of the reservoir
model.
The EMpower software design is predicated on the need for
parallel processing in order to take full advantage of recent
advances in computer architecture. Ensuring that this software
design is also compatible with unstructured gridding and
localized modeling required the development of a complex but
efficient data management infrastructure. As a design
principle, the parallel computational algorithms do not impose
restrictions on the selection of the grid employed.
EMpower provides users with an easy-to-use and
comprehensive graphical interface to build, manipulate, and
maintain reservoir simulation model input data and results.
“Visual Facility Management” allows users to build facility
management strategies and to define facility networks
graphically and interactively. Constraints and actions for
wells and facilities are implemented using a graphical
programming interface that results in logical flow charts
(“logic diagrams”) of the facility management logic. 3D
visualization of simulation input data and results is available
from within the GUI. A comprehensive embedded on-line
help and tutorial system assists novice users in getting started
with simulation and provides expert users the ability to review
less-frequently-used features as needed. Users also are able to
monitor the progress of simulation runs in real time using
interactive simulation monitoring tools. A batch queuing
SPE 68116 EMPOWER : NEW RESERVOIR SIMULATION SYSTEM
system is built into the application to allow management of
multiple jobs.
Other EMpower features include integrated surface
networks, special fluid models for water vaporization, various
formulations for miscible and near-miscible displacements,
rock compressibility options, multiple well options including
non-Darcy flow, integrated upscaling including optimal layer
selection, and feedback (use of simulation results to improve
the geologic model).
Unstructured Grids
A key element of the new simulator is the use of unstructured
grids. Breaking the mold of using rectangular or regular grid
systems enables more accurate and detailed representation of
complex geologic and engineering features such as faults,
pinchouts, fluid contacts, and horizontal or multi-lateral wells.
The simulator is devoid of any addressing or computational
concepts founded upon the conventional “I-J-K addressing.”
This fundamental departure influences code design, reservoir
engineers’ preparation of input data, and design and
construction of simulation models. The code design adopted
to accommodate unstructured grids also provides a natural
framework for the integration of well, facility and reservoir
calculations for increased computational stability. EMpower
readily accommodates models with multiple fields where the
fields are individually gridded from separate geologic models
and combined into a single simulation model.
Figures 1(a) and 1(b) are images from a simulation model
that exploits many of the features discussed in the previous
paragraph. Three reservoirs are included in the simulation
model and each was constructed from a separate and distinct
geologic model.
Unstructured gridding in EMpower allows a fit-for-purpose
discretization of each reservoir adapted to the specific
situation being modeled. This approach avoids the work-
arounds needed to force-fit a grid within the constraints of a
hierarchically structured Cartesian gridding system of the kind
used by conventional simulators. Unstructured gridding
provides a more natural framework for modeling complex
wells, e.g. slanted, horizontal, and multi-lateral wells, while
honoring the geology. Gridding to incorporate complex faults
and pinchouts, though inherently complex, becomes tractable
using unstructured grids. Reservoir A (on the left in
Figure 1(b)) is an example of a meandering channel sand.
Horizontal wells are placed along channel as shown in
Figure 2. Finer grids are used along the channel sand and in
the local vicinity of the horizontal wells. Extensive use of the
ability to grid to faults was employed for Reservoir C, as
illustrated in Figure 3.
Using a finer grid only where needed employs
computational resources most effectively. For detailed
understanding of behavior in a particular region, localized grid
refinement provides flow details without impacting the model
in other connected regions. Grids in areas of lesser interest are
easily coarsened.
As is well known, flexibility in gridding typically comes at
the expense of some additional effort in the solution method.
3
However, in practice we have found there are efficiencies and
simplifications associated with areal grid refinement, handling
of pinch-outs, and the elimination of inactive cells, that can
more than offset the increased work of an unstructured solver.
Practical examples of areal grid refinement can be observed in
Figures 2 and 3. Figure 4 illustrates the use of pinched-out
layers. It is expected that as solution methods for
unstructured grids continue to improve and the complexity of
geologic models continues to increase, unstructured gridding
will become even more advantageous.
Upscaling, Optimal Layering, and Feedback
The simulation system employs Geoquest's FloGridTM
Software as the primary gridding package. The geologic
model can be imported into this software in a variety of
formats, including map-based models and the industry-
standard “RESCUE” format. Because full-field geologic
models easily contain tens or hundreds of millions of cells, it
is more cost-effective and practical to intelligently upscale
these models to an accurate unstructured grid than to create
huge, rectangularly-gridded models with no upscaling. The
gridding package has been modified by ExxonMobil to
include EMpower-proprietary methods for optimal layering and
flow-based scale averaging, in addition to the basic
capabilities in the vendor software. Both of these proprietary
capabilities have been developed for the unstructured grid
used within EMpower.
The package for optimal layer selection uses objective
functions to estimate the error in simulation results associated
with combining multiple layers from the geologic model into
coarser simulation layers10. This module automatically and
optimally defines which layers in the geologic model are to be
combined into the coarser simulation layers. This capability
also gives the user an idea of how many layers are needed to
yield a prescribed accuracy.
Single-phase reservoir simulations are performed on grids
that are much finer than the simulation model to determine the
flow-based properties. For this purpose the gridding package
has been modified to generate “fine-grid” models specifically
for scale averaging. Figure 5 illustrates the fine-grid geometry
typically used for flow-based scale averaging. The actual and
fine-grid simulation grids are both exported from the gridding
package into the simulator. Flow-based scale averaging
methods are well established for structured grids with brick-
type blocks. Significant development was required to extend
these methods accurately to unstructured grids employing a
wider variety of block shapes. A proprietary, internally-
developed approach is used for upscaling on unstructured
grids.
Algorithms for optimal layering and flow-based scale-up
for unstructured grids are tightly integrated in the EMpower
system. Previously, specialists were almost exclusively
responsible for generating the grid and scaled-up properties
delivered as a product to the reservoir engineer. This process
inherently limited the extent to which a reservoir engineer
experimented with different grids or modified the grid based
on interim simulation results. Integrating the scale averaging
4 B.L. BECKNER, J.M. HUTFILZ, M.B. RAY, J.F. TOMICH
technologies into the EMpower desktop environment enables the
reservoir engineer to have full flexibility to modify the grid
and then quickly and efficiently scale-up the properties for the
new model.
EMpower has feedback capabilities to help engineers and
geoscientists use additional knowledge gained from field
performance history to improve the accuracy of both geologic
and simulation models. These capabilities, essential
components of the feedback process, make it possible to
maintain reservoir models that honor both initial reservoir
description data (including geologic interpretations) and field
performance history. Here, field performance history includes
pre-development well tests as well as post-development
production performance. Changes made to the simulation
model to match field performance history are fed back as
additional conditioning data to the geologic model, thereby
updating the geologic model with production-generated
reservoir characterization information. Feedback capabilities
in EMpower make it possible to import geologic models, co-
visualize geologic and simulation models, generate statistics
on geologic and simulation model properties, link geologic
model cells to simulation nodes, edit model properties guided
by geologic constraints, and export conditioning data to the
geologic model.
Surface Networks and Multi-Field
The mathematical model in EMpower for subsurface flow which
uses an array of pressure nodes and the connections between
those nodes, allows the user to selectively include as an
integral part of the simulation some or all of the production
and injection facilities (e.g. wells, tubing, surface flowlines,
manifolds, separators). An integrated reservoir and surface
facility modeling capability is attractive for two reasons. First,
it allows more accurate modeling of the entire
reservoir/facility system. Second, this implicit solution option
provides improved stability by circumventing problems
associated with rapid pressure and saturation changes near
wells that are often encountered in conventional reservoir
modeling approaches where surface facility modeling is either
absent or entirely explicit.
This capability allows all the facility components shown in
Figure 1, wellbore to surface, flowlines and riser, to be
included in the computational model. For multiple fields the
interaction between facility rates and pressures is implicitly
solved at each time step. This capability is also important for
multi-lateral well modeling, as illustrated by Figure 6, where
pressure drops in wellbores and the interaction between well
branches are important.
EMpower provides a graphical interface for defining the
integrated facility network. This interface displays the facility
network as a collection of icons attached to one another via
connection arrows. An example of a facility network diagram
for the three-field model of Figure 1 is shown in Figure 7(a).
Different icons are used to represent distinct facility types
such as wells, separators, network junctions and manifolds.
The connection arrows represent flowlines between the “node-
SPE 68116
like” facilities. Figure 7(b) shows an alternate view of the
three-field facility network scaled to reservoir dimensions.
The flow rates and pressure drops within the surface
facilities can be accurately modeled if the user chooses to
integrate more of the facility network into the calculation
model. Facility management is fully integrated with the
facility network layout so that each node or connection in the
facility network can be queried and modified in various ways
during the simulation.
The EMpower facility network data design was developed to
maximize flexibility and extensibility so that facility types can
be extended and new facility types added without requiring
extensive software changes. Facilities are defined
“generically” and contain “attributes.” The attributes and
much of the specialization for given facility types are defined
in a data file, not in compiled code. With this design,
extensions to the facility network functionality, such as adding
new facility attributes or defining new facility types, can be
made simply by changing the data file.
EMpower supports import of a variety of facility network
data from external sources. Historical or time-varying data,
including well perforation intervals, production rates, injection
rates and pressure measurements, can be imported from text-
delimited files. Such data are automatically converted to the
required boundary conditions in the simulation model, saving
the user time and effort when constructing a simulation model
or developing sensitivity cases.
A natural extension of the integrated subsurface/surface
model is the inclusion of a multi-field capability. EMpower’s
multi-field simulation capability emphasizes ease-of-use.
With just a few clicks of the mouse, individual field
simulation models can be merged into a multi-field simulation
model. EMpower automatically handles the time-consuming
tasks of creating a uniform timeline, merging input data,
facility networks and facility management logic, and dealing
with duplicate names in regions, wells, facilities, etc. The
multi-field model is ready to run as soon as the engineer adds
the facility network components (that connect the fields
together) and updates the field-level rates and constraints.
Facility Management Logic
Another significant aspect of EMpower is the use of
programmable facility management. ExxonMobil has
advocated this approach to simulator facility control for many
years11. One goal of the new reservoir simulator is to allow all
reservoir engineers to have extensive capabilities for
constructing, sharing and customizing facility management
routines. This has led to the introduction of a "Visual Facility
Management" capability illustrated in Figure 8. "Visual
Facility Management" within the user interface comprises two
views: facility network view and logic diagram view. The
logic diagram view serves as a way for users to develop and
execute Facility Management Logic (FML). FML is a
customized sequence of steps used to monitor and control the
behavior of facilities throughout the simulation run. The FML
functionality is flexible and powerful enough to enable an
engineer to model and solve complex reservoir development
SPE 68116 EMPOWER : NEW RESERVOIR SIMULATION SYSTEM
schemes, yet has a convenient graphical user interface that is
both intuitive and informative.
When the user is ready to launch a simulation, EMpower
converts the logic into standard C++ code. This C++ is sent to
the simulator as part of the simulation input file, where it is
compiled and dynamically linked into the simulator at run
time.
Graphical User Interface and Interactive Simulation
The EMpower graphical user interface is an all-in-one integrated
product for ease of use. Its capabilities include gridding, up-
scaling, 3D visualization, data entry, saturation function
generation, PVT table generation, visual facility network,
visual facility management logic, job scheduling, and results
analysis. A comprehensive help system includes fly-by tool-
tips, a context-sensitive documentation window, ExxonMobil
reservoir engineering and simulation guidelines, and links to a
complete text-based or voice-based tutorial. Figure 9 shows a
view of the EMpower GUI.
The GUI was designed with flexibility and maintainability
in mind. It is based on a complete class hierarchy of generic
data types in combination with rule-based data entry. This
design allows most simulator changes to be quickly reflected
in the GUI by simply changing a data file, then running a
utility program against the EMpower Database, with no re-
compilation necessary. This data file defines all possible
simulation input data and rules that define the dependencies
between them. For example, this file specifies the
dependencies between the phases/components to be modeled,
the required columns in PVT tables, and the required set of
saturation function (relative permeability and capillary
pressure) tables. The rule system constantly reevaluates user
input so that required versus optional inputs change "on the
fly." This design makes it much easier to keep the GUI
synchronized with the simulator.
An EMpower simulation can be scheduled for batch
execution or can be monitored interactively in real-time. In
interactive mode, the GUI launches the simulator, then
"connects" and communicates directly with the running
simulator. While connected, the user can track progress, load
and review intermediate results, pause, continue, or cancel the
simulation at any time.
Multi-User Database and Data Management
All of the EMpower input data are stored as objects in an object-
oriented database (not a file-based system). Engineers are
freed from having to keep track of simulation input files, and
multiple users can log on and see the same data. This allows
engineers in different physical locations to collaborate in real
time. Robust disaster recovery features prevent significant
data loss in the event of software, hardware or power-system
failures.
EMpower's sophisticated database schema allows sharing of
input data among cases. Increasing the shared component of
data among cases reduces proliferation of copied data,
improves data consistency, and reduces the total disk space
required. This is a powerful tool for sensitivity analysis since
5
changes may be made private (i.e. not shared with other cases)
or may be propagated to many cases via "shared data."
An internal security scheme is implemented using internal
user ID's, groups and permissions, thereby allowing model
owners to selectively grant "read-only" or "update" rights to
other engineers.
The database and data-sharing infrastructure of EMpower is
leveraged to facilitate progressive development and concurrent
application of individual and multi-field models. Rather than
combining individual grids from stand-alone field models into
a larger grid, the multi-field model simply "shares" these items
with the individual models. As the study progresses, the
engineer or geoscientist can continue to run each individual
field stand-alone or in the context of the multi-field model.
Changes made in either context are automatically reflected in
the other.
Testing, Training and Application
Reservoir simulation is a critical tool that ExxonMobil
employs in making multi-billion dollar investment decisions.
An extensive test program was implemented for EMpower that
ranged from lower level programmer tests to tests employing a
series of full-field models developed by reservoir engineers.
These models were selected to cover a broad spectrum of
technical features in the simulator and to be representative of
the current business needs. Full-field test model results using
EMpower were compared and benchmarked against results for
very similar models constructed using the company's previous
simulators. At the completion of testing for Version 1.0 of the
simulator, a series of one week hands-on training sessions
were initiated that have now been attended by close to 200
users. The new simulator is the preferred simulation tool for
all new reservoir simulation studies. Unstructured gridding is
the new feature having the widest initial application and
arguably has been the single greatest learning challenge for
experienced simulation users.
A suite of full-field model studies initially was used to test
the simulator and provide confidence in its use for business
decisions. For simple models with consistent rock and fluid
properties the simulator produces results numerically identical
to those obtained with the previous generation of simulators.
However, the new simulator was not designed merely to
produce identical results; new formulations are employed
wherever they are technically justified.
When rectangular grid blocks are used and special
measures are taken to ensure the rock properties are closely
matched, full-field models produce nearly identical results to
previously developed simulators. Figure 10 shows oil, water
and gas production for a full-model with 221,000 rectangular
grid blocks. Very similar results are also shown for a case
using an unstructured grid that employs only 85,000 grid
blocks. In the unstructured grid case, finer gridding is
employed around wells located in an oil rim. The model also
significantly reduced the number of grid blocks by pinching
out numerous layers.
The model shown in Figure 11 employs an unstructured
grid with complex non-vertical faulting. Deviating grids to
6 B.L. BECKNER, J.M. HUTFILZ, M.B. RAY, J.F. TOMICH
conform to faults, while possible, is known to introduce
numerical inaccuracies. Instead vertically stair-stepped faults
were used to model the highly deviated faults. Prior to
pinching out blocks, the initial grid consisted of 1.6 million
blocks. However, after pinching out blocks, the model
reduced to only about 100,000 blocks. In this case,
unstructured gridding enabled simulation of a model that
incorporates as much accurate geologic features as a finely
gridded Cartestian grid with significantly fewer gridblocks.
The simulator also is proving to significantly reduce
simulation cycle-time through ease-of-use and integration with
geologic models. For example, one successful application
includes history matching and interpretation of well testing
results where several different fault interpretations were
investigated. Another application consists of a series of
reservoir studies of a large carbonate accumulation where
alternative development plans are being explored. In this case,
simulation models have been developed for single wells, five-
spot patterns, multi-lateral wells, and full-field development
plans using both black oil and compositional formulations, all
based on data extracted from a single underlying full-field
geologic model. In this study, the shared database allowed
rapid development of this suite of models.
Conclusion
This paper highlights a number of distinguishing technologies
including computations made on a native, unstructured grids,
integration of facilities and models for multiple fields, and
highly customizable facility/well management logic to give
our engineers and geoscientist the technologies necessary for
best-in-class asset management. EMpower is now the standard
reservoir simulator for ExxonMobil. ExxonMobil will
continue to develop reservoir simulation technologies because
they are key to delivering maximum full-life value from our
assets.
In a world where reservoir simulators are readily available,
what have we gained by this development? EMpower is a next-
generation reservoir simulator that embodies ExxonMobil’s
approach to state-of-the-art reservoir simulation technology
and contains features needed to best manage ExxonMobil's
asset base. The management of our major assets requires a
balanced technology approach -- one that focuses not only on
the use of technology and the tuning of that technology to the
specific assets, but that also anticipates and accounts for the
advancement of technology, both in terms of the
science/algorithms and computing infrastructure. As just one
example, considerable corporate experience in condensate
reservoirs has lead to specific formulations and models
developed on the basis of many years of laboratory and field
data that, in our opinion, represent best technology practices in
condensate field development. The "physics" and physical
models corresponding to many of these technical findings
have been implemented in our reservoir simulators. Viewed
in this light, the simulator within ExxonMobil is far more than
a commodity application. It is the vehicle for the delivery of
proprietary technologies of particular relevance to our assets
that have been or will developed in the future. Thus EMpower
SPE 68116
incorporates our best practices and preferred
technologies/models. In short, it is a cornerstone of our
reservoir knowledge management.
Acknowledgements
The authors wish to acknowledge A.D. Banki, T.J. Boone,
K.T. Guenther, R.T. Mifflin, S.C. Netemeyer, S.C. Niranjan,
R.E. Pugh and S.K. Verma for their valuable assistance to this
paper. The authors also acknowledge the management of
ExxonMobil Upstream Research Company for permission to
publish this paper.
References
1. Kendall, R.P., Morrell, G.O., Peaceman, D.W., Silliman, W.J.
and Watts, J.W.: "Development of a Multiple Application
Reservoir Simulator for use on a Vector Computer," presented
at the Middle East Oil Technical Conference of the Society of
Petroleum Engineers, Bahrain, March 14-17, 1983.
2. Harper, J.L., Heinmann, R.F., Ray, M.B. and Stephenson, P.E.:
“A Compositional Simulator For Performing Large Field
Studies in a Vector Computing Environment,” SPE 13714,
presented at the 4th SPE Middle East Oil Technical Conference,
Bahrain, March 11-14, 1985, 279-290.
3. Watts, J.W.: “A Compositional Formulation of the Pressure and
Saturation Equations,” SPE Reservoir Engineering, vol. 1, no. 3.
May 1986, 243-252.
4. El-Mandouh, M.S., Bette, S., Heinemann, R.F., Ogiamien, E.B.,
Bhatia, S.K.: “Full-Field Compositional Simulator of Reservoirs
With Complex Phase Behavior,” SPE 25249, presented at the
12th SPE Reservoir Simulation Symposium, New Orleans,
Louisiana, February 28 – March 3, 1993, 167-186.
5. Beckner, B.L., Chan, H.M., McDonald, A.E., Wooten, S.O. and
Jones, T.A.: “Simulating Naturally Fractured Reservoirs Using a
Subdomain Method,” SPE 21241, presented at the 11th SPE
Reservoir Simulation Symposium, Anaheim, California,
Feburaruy 17-20, 1991, 383-396.
6. Mifflin, R.T., Watts, J.W. and Weiser, A.: “A Fully Coupled,
Fully Implicit Reservoir Simulator For Thermal and Other
Complex Reservoir Processes,” SPE 21252, presented at the 11th
SPE Reservoir Simulation Symposium, Anaheim, California,
Feburaruy 17-20, 1991, 457-470.
7. Hienemann, Z.E.: "Advances in Gridding Techniques,"
presented at the Fifth International Forum on Reservoir
Simulation, Muscat, Oman, December 10-14, 1994.
8. Verma, S., and Aziz, K.: "A Control Volume Scheme for
Flexible Grids in Reservoir Simulation," SPE 37999, presented
at 1997 Reservoir Simulation Symposium, Dallas, Texas, June
8-11 1997, 215-227
9. Gunasekera, D.L., Cox J., and Lindsey, P.: "The Generation and
Application of K-Orthogonal Grid Systems," SPE 37998,
presented at 1997 Reservoir Simulation Symposium, Dallas,
Texas, June 8-11 1997, 199-214
10. Stern, D.A., and Dawson, A.: "A Technique for Generating
Reservoir Simulation Grids to Preserve Geologic
Heterogeneity," SPE 51942, presented at 1999 Reservoir
Simulation Symposium, Houston, Texas, Feburary 14-17, 1999,
487-500.
11. Miertschin, J.W., and Weiser, A.: "A Flexible Approach to
Predictive Well Management Via User-Defined Strategies,"
SPE 19848, presented at 64th Annual Technical Conference and
Exhibition of the SPE, San Antonio, Texas, October 8-11, 1989,
817-826.
SPE 68116 EMPOWER : NEW RESERVOIR SIMULATION SYSTEM 7

Figure 1(a). Image of a multi-field reservoir simulation model.

Three reservoirs constructed from different geologic models are
combined into one simulation model with common well
management and production facilities. Colors identify regions of
initial oil saturation in the three reservoirs.

Figure 1(b). Image of the porosity distribution in the three

reservoirs.
8 B.L. BECKNER, J.M. HUTFILZ, M.B. RAY, J.F. TOMICH SPE 68116

Coarse Gridding

Finer Gridding

Local Grid
Refinement at
Horizontal Wells

Figure 2. Example of the unstructured gridding used for

Reservoir A, where finer grids are placed in the higher porosity
sands, and coarse grids in the lower porosity sands. Grid
refinement is employed around a series of horizontal wells along
the channel sands. The color scale represents porosity.

Gridding to
Faults

Figure 3. Example of unstructured gridding from Reservoir C

where the engineer has gridded to a series of faults.
SPE 68116 EMPOWER : NEW RESERVOIR SIMULATION SYSTEM 9

Pinched-Out Layers

Figure 4. Example of pinched-out layers from a channel sand in

Reservoir C.

Finer Grid used for Flow-

Based Scale Averaging Simulation Grid

Figure 5. Illustration of an unstructured hexagonal grid used for

simulation, compared with a finer unstructured grid used for flow-
based scale averagibg to evaluate properties used in the actual
simulation model.
10 B.L. BECKNER, J.M. HUTFILZ, M.B. RAY, J.F. TOMICH SPE 68116

Location of
Multi-lateral
Well

Laterals

Figure 6. Images of a multi-lateral well embedded in Reservoir A.

The grid has been cut away to expose the well and its location
relative to geologic layering in the model.
SPE 68116 EMPOWER : NEW RESERVOIR SIMULATION SYSTEM 11

Subsea Manifold Common Boundary Node

Flow Line Riser

Reservoir B Reservoir C

Wells

Reservoir A

Figure 7(a). Graphical interface view of the facility network for the
multi-field model. Note that the entire facility network below the
common boundary node can be included within the
computational bounds of the reservoir simulation.

Reservoir A Reservoir B Reservoir C

Figure 7(b). The three-field facility network diagram scaled to

reservoir orientations.
12 B.L. BECKNER, J.M. HUTFILZ, M.B. RAY, J.F. TOMICH SPE 68116

Figure 8. Example of an "operation" used to input visual well

management for controlling the reservoir simulation.

Figure 9. Graphical User Interface for the EMpower simulator.

SPE 68116 EMPOWER : NEW RESERVOIR SIMULATION SYSTEM 13
Rate (MSTB/D, MMCF/D)

240
MARS
Mars
200 EMpower Rectangular Grid
EM rect Gas
EMpower Hexagonal Grid
EM var-hex
160 EMpower Hex & Pinchouts
EM var-hex, pinchout
120
80 Oil

40 Water
0
88 89 90 91 92 93
Year
Figure 10. Plot comparing results from different models: (1)
previous generation simulation with 221,000 blocks; (2) EMpower
using a rectangular grid with 221,000 blocks; (3) EMpower using an
unstructured, hexagonal grid with 139,000 grid blocks; and (4)
EMpower using an unstructured, hexagonal grid and layer pinch-
outs reducing the model to 87,000 grid blocks.

Figure 11. Image of a reservoir with complex faulting where

unstructured gridding enabled a much more complex geological
model to be employed relative to a traditional structured gridding
approach.