PORE-SCALE MODELING:
From micron-scale imaging
to field-scale consequences

FRACTURED RESERVOIR MODELING AND SIMULATION

RESERVOIR SIMULATION SYSTEM FOR THE NEXT DECADES

3D and 4D Seismic and its Role in Field Development and Optimization

RULES FOR AN OPTIMAL FLOW SIMULATION GRID GEOMODELING AND FLOW SIMULATION: THE NECESSARY INTEGRATION

Ivar Aavatsmark,
Zoltán E. Heinemann

SOLVING RESERVOIR SIMULATION EQUATIONS

MODERN TECHNIQUES FOR HISTORY MATCHING

STREAMLINE BASED HISTORY MATCHING AND PREDICTION

Sunday, December 9th, 2007

KEYNOTEADDRESS:
‘ADNOC’

presented by Mohamed Juma Abdulla Bin Juma

Manager, Onshore DivisionE&P Directorate

Abu Dhabi National Oil Co. (ADNOC)

Mohamed Juma Abdulla Bin Juma graduated from University of Tulsa in Petroleum Engineering – 1981. Immediately after graduation joined ADCO, during which he worked and gained experience in drilling, production and reservoir engineering (both measurement and simulation).
Key positions held in ADCO were Head of Reservoir Engineering, Field Development Manager and Petroleum Development Manager. He was transferred to ADNOC in year 2000 as Onshore Division Manager. Mr. Juma current responsibility as Onshore Division Manager is focused on managing development, production and commercial issues relating to oil, gas and condensate within ADCO concession. He also manages National Drilling Company (NDC) activities on behalf of ADNOC. He is a Board Member in NDC and Chairman of Board Advisory Committee of NDC. He is also Board Member in GASCO. Mr. Juma represents ADNOC in all Technical Committee Meetings of Onshore OPCOs.

Mohamed Juma Abdulla Bin Juma

Bio-Data

Graduated from University of Tulsa in Petroleum Engineering - 1981.

Immediately after graduation he joined ADCO, where he worked and gained experience in drilling, production and reservoir engineering (both measurement and simulation).

Key positions held in ADCO were Head of Reservoir Engineering, Field Development Manager and Petroleum Development Manager. He was transferred to ADNOC in year 2000 as Onshore Division Manager.

Mr. Juma current responsibility as Onshore Division Manager is focused on managing development, production and commercial issues relating to oil, gas and condensate within ADCO concession. He also manages National Drilling Company (NDC) activities on behalf of ADNOC.

He is a Board Member in NDC and Chairman of Board Advisory Committee of NDC. He is also Board Member in GASCO.

Mr. Juma represents ADNOC in all Technical Committee Meetings of Onshore OPCOs.

Sunday, December 9th, 2007

Pore-Scale Modeling: from Micron-Scale Imaging to Field-Scale Consequences

presented by Martin J. Blunt
Department of Earth Science and Engineering
Imperial College London

The last ten years witnessed significant advances in our understanding of the pore-scale mechanisms of multiphase displacement, thanks to the ability to image the pore space at resolutions of a few microns using micro-CT scanning and increasingly sophisticated numerical modeling techniques. An overview of these methods and some of the major results are presented. It is now possible to make predictions of multiphase flow properties based on an accurate characterization of the geometry of the pore space and the wettability. Examples of this approach are given for single-phase transport, non-Newtonian flow, two-phase flow and three-phase displacements involving different sequences of water and gas injection. This work helps us to interpret experimental measurements of relative permeability and capillary pressure and to predict these quantities where data is lacking - such as when there are structural or wettability variations across the reservoir, or where three phases are present.

Pore-scale processes are important in field-scale transport; they are not somehow swamped by larger heterogeneity in permeability. Both the pore-scale physics and the geological structure of a reservoir have a first-order impact on recovery. This is illustrated by a series of case studies, including the study of dispersion in tracer flow, waterflooding a reservoir with a wettability transition above the oil/water contact and the design of carbon dioxide storage in aquifers and mature oil reservoirs.

Martin J. Blunt

Martin Blunt is head of the Department of Earth Science and Engineering at Imperial College London. He joined Imperial in June 1999 as a Professor of Petroleum Engineering. Previous to this he was Associate Professor of Petroleum Engineering at Stanford University in California. Before joining Stanford in 1992, he was a research reservoir engineer with BP in Sunbury-on-Thames. He holds MA and PhD (1988) degrees in theoretical physics from Cambridge University.

Professor Blunt's research interests are in multiphase flow in porous media with applications to oil and gas recovery, contaminant transport and clean-up in polluted aquifers and geological carbon storage. He performs experimental, theoretical and numerical research into many aspects of flow and transport in porous systems, including pore-scale modelling of displacement processes, and large-scale simulation using streamline-based methods. He has written over 100 scientific papers and is on the editorial boards of three international journals.

Sunday, December 9th, 2007

FRACTURED RESERVOIR MODELING AND SIMULATION

presented by Jean Burrus
Chairman of the Management Board of Beicip-Franlab

Fractured reservoirs are not easy to model for the geologist and for the reservoir engineer. The main problem encountered is that fractures are seldom directly observed on cores or logs, and when fracture occurrences are found by geologists, reservoir engineers usually do not know how to use this information to develop more accurate reservoir flow models. The recently developed DFN (discrete fracture network) approach associated with more accurate physical simulation of matrix-fracture exchange have provided a powerful way to integrate the static and the dynamic understanding of fractures. We will show through examples of studies of fractured fields in the Middle East how state-of-the art techniques of static modelling and dynamic simulation of fractured reservoirs help to renew our understanding of fractured reservoirs. Our examples will be based on the new generation FracaFlow and PumaFlow tools developed by IFP with the support of the industry. The impact on the management of these fields will be shown.

We will discuss the following aspects. How can we know a reservoir acts as a fractured reservoir? What is the fracture scale to be considered? How does fracture spacing impact production mechanisms? How to develop stochastic fracture models from available data? How to predict fracture permeability? How to upscale fracture permeability? When should single medium versus dual medium models be preferred?
We will stress the outmost importance of an appropriate physical description of matrix-fracture exchange terms in predicting correctly production from fractured reservoirs.

Jean Burrus

Jean Burrus is the Chairman of the Management Board of Beicip-Franlab. Jean graduated from the Paris School of Mines in 1980 and holds a Doctorate degree in Quantitative Hydrogeology of Paris School of Mines. After a teaching experience in 1980-1982 at the Algerian Petroleum Institute, he joined IFP Geology Geochemistry Division in 1982. His research focused on hydrocarbon system modeling, basin dynamics, oil and gas generation and migration, and quantitive reservoir modeling. He held various management positions in IFP in Geoscience and Reservoir Engineering Divisions. In 1997, Jean joined Beicip-Franlab as Technical Director and Head of the Software Division. Jean is the recipient of the Barbier Prize of the French Geological Society. He is an active member of several professional societies, including AAPG, EAGE and SPE.

Sunday, December 9th, 2007

RESERVOIR SIMULATION SYSTEM FOR THE NEXT DECADES

presented by Long Nghiem
Vice-president
Research and Development, with Computer Modelling Group Ltd.
and
Jeroen C. Vink
principal reservoir engineer and research scientist
Shell International Exploration and Production B.V.

Reservoir simulation technologies, workflows and data handling requirements have evolved rapidly over the past two decades. At the same time, hardware performance and characteristics have also progressed by leaps and bounds. The reservoir simulation software developers have been faced with the challenge of working with different pieces of legacy software that are often not fully integrated and that are written in programming languages that do not embrace modern software engineering concepts (e.g. object oriented programming, embedded scripting, etc.). The original solution methods were not designed to effectively handle new recovery processes, new gridding techniques and the demand for coupled reservoir flow, geomechanics, wellbore and surface facilities. Codes need to be frequently rewritten in attempting to take advantage of the rapid evolution of computer hardware. Trying to implement such changes in existing data structures has many drawbacks that give rise to duplication of data and other inefficiencies across the system. From the user viewpoint, the simulation components are not as easy to operate as is required in fully integrated workflows, and a lot of time is spent in inefficient manual procedures to find work-arounds for the limitations of the system. Although, the software developers are often able to satisfy the user's demands by making clever fixes to the existing codes, it has become evident that a redesign of the whole simulation system is necessary to meet the current and future requirements of the reservoir engineers.

This paper addresses features and requirements of a new reservoir simulation system that will meet the challenges of the next few decades. The topics to be discussed include:

• Seamless Workflow Integration covering the full modeling chain from "pore throat to point of sale", multiscale scenarios, and plug-ins for 3rd party software.

• Powerful Graphical User Interface that empowers the novice reservoir engineer and unleashes the full power and creativity of experts.

• Data Trail Audit to keep track of the differences in data and results between a multitude of runs.

• Assisted History Matching, Optimization and Uncertainty Assessment to help the engineer get the most meaningful information out of a study with uncertain data and make the proper decisions.

• Speed and Efficiency for ultra large and complex simulation models (10^7 gridblocks) on distributed multi-threaded computing environments with efficient gridding and multiscale techniques.

• Advanced Physics and Capabilities including a unified approach for black-oil, compositional, thermal and chemical simulation, the ability to perform full field simulation of thermal, EOR or unconventional processes with multiscale physics and the potential to analyze recovery mechanisms with coupling of fluid flow and geomechanics.

• Coupling to Comprehensive Models for Wellbore and Surface Facilities for modelling multilateral and smart wells, optimal control for wells and facilities and full field life-cycle optimization with surface facilities.

• Flexible Nonlinear Equation Solution Methods that include IMPES, IMPSAT, Fully Implicit formulations, inner and outer iterations for flash calculations and variable degrees of coupling with surface facilities and geomechanics.

• Unstructured, dynamic, grids that provide the resolution needed to model complex process physics, only where needed.
   

Long Nghiem

Long Nghiem is Vice-president, Research and Development, with Computer Modelling Group Ltd. He joined the firm in 1977 and has been involved in the research, development and application of reservoir simulation technologies. He has authored over 70 papers on various aspects of reservoir simulation. His current interests include EOR and CO2 processes, greenhouse gas storage, next generation simulator development, history matching, optimization and uncertainty assessment. He holds a Ph.D. degree in Petroleum Engineering from the University of Alberta, and degrees in Chemical Engineering from the University of Waterloo and the University of Montreal. He is also an Adjunct Professor in the Chemical and Petroleum Engineering department at the University of Calgary.

Jeroen C. Vink

Jeroen C. Vink is principal reservoir engineer and research scientist in Shell International Exploration and Production B.V. He has a PhD in theoretical high-energy physics from the University of Amsterdam. After a period of academic research in Germany and at the University of California at San Diego, he joined Shell in 1994. Here he started working on Shell's in-house reservoir simulator MoReS. He has worked on all aspects of flow simulation and upscaling, with a focus on improving (linear) solver techniques and parallelization. Currently he is working on designing and implementing "Next Generation" simulation techniques in collaboration with the Computer Modeling Group Ltd. His interest focuses on massively parallel (linear) solvers, multi-scale flow methods, integrated uncertainty handling and new techniques for simulating enhanced oil recovery processes.

Monday, December 10th, 2007

3D and 4D Seismic and its Role in Field Development and Optimization

presented by Lars Sonneland
Schlumberger Stavanger Research

Introduction

A new approach towards model building with the promise of significantly shortening the turnaround time of 3D model building from seismic is presented. By introducing a unifying framework, efficient representation of models throughout the lifecycle of a reservoir is enabled, all the way from velocity to simulation models.

The industry is demanding methods that can coop with increasingly complex reservoirs and geometries that cannot practically be represented by bounding surfaces, such as top and base horizons (e.g., carbonates, channels, salt...) . These demands necessitates a new generation of volume interpretation tools. The truly three-dimensional nature of seismic data enables model-building using volumes instead of primitive surfaces more attractive and convenient. Volume interpretation also provides better handling of geometries induced by volume properties that are not directly represented in a surface framework.

An important requirement of any new interpretation tool is that it must be fast and intuitive to use. Making informed decisions necessitates good data representation, analysis and interaction capabilities, thus, putting high focus on user interface and interaction capabilities. An excellent approach to a more user friendly, high level and informative interpretation system is seen in the recently introduced interpretation paradigm of system level interpretation . Here the key idea is to automate the task of extracting low-level primitives, such as horizons and faults, and present the user with an overall system level overview of these pre-generated primitives. We will define the set of model primitives a model repository. This represents a top down approach, where the users starts with the big picture, and zooms in on the smaller details, instead of spending time building the details and maybe never see the bigger picture. The user can then spend time analyzing the data, applying and developing geological knowledge important for the field. This approach to interpretation is orders of magnitudes faster than conventional approaches, and in addition offers far superior data information and analysis capabilities.

Geosteering using model repositories
In exploration, appraisal or early production scenarios, scarce or little information about the reservoir is known; uncertainty in reservoir heterogeneity is large and might impact the drilling success. As the well is drilled, new information becomes available. Utilizing this information to obtain an improved understanding of the geometry of the reservoir is desirable for optimal well placement. The well path can then be steered relative to the updated model and thereby maximize productive length and the production profile, i.e., precisely place the well relative to fluid contacts.

Initial model repository
The initial model in our workflow is constructed from a set of horizon patches and fault patches extracted from the 3D model repository. One horizon patch represents an interface in 3D with a similar reflectivity ( or similar seismic attribute response).. A sequence of horizon patches defines a geobody class. Such classes are generated by classification .

Real time updating of reservoir models
These geobody classes need to be calibrated with well- logs to identify which lithology they represent . By designing the appropriate LWD log-suite real-time calibration of the geobody landscape is enabled . The LWD data is provided continuously through a real-time link using InterACT. These log-suites are first inverted log-sample by log-sample to lithology in a probabilistic manner . In order to calibrate the geobodies with the accumulated properties a proper upscaling is required. The logs are on centimeter scale and the geobodies are in meters. Hence, there will be many log samples within one geobody. The probable lithology for a geobody is determined by finding all the log samples that fall within the extent of the geobody, and use their aggregated probabilities to decide the most probable lithology. The probabilistic lithology for the geobody is updated and similarly the other members of the same geobody class. This creates a probabilistic lithology landscape of geobodies which enables look-around view . This landscape is continuously updated as we receive new LWD data.
A prerequisite for calibrating the geobody model with well logs is proper time-to-depth conversion, so that the log responses can be mapped into the correct positions in the model. By using e.g. seismic while drilling measurements or periscope observations, the 3D time-to-depth conversion is continuously refined by adding new control-points to the velocity model. The 3D velocity refinement is applied to the whole model repository and all historical control-points are honored .

Conclusion
Real time generation of a probability landscape using LWD data provides crucial information for optimal placement of wells during drilling through look ahead.
Planned well trajectories can then be modified according to the updated probability landscape to optimize well production.

Lars Sonneland

Lars Sonneland is Research Director in Schlumberger. He joined Schlumberger after a post - doc position in applied mathematics at the University of Bergen ,Norway . Lars has managed several of the companys research programs from their centers in Ridgefield, Connecticut , Cambridge , UK and Stavanger,Norway. Recipient of the Association of Chartered Engineers Technical Award ,the Geophysical Award and the Schlumberger "Best R&D Project Award" , Lars has played a major role in the development of 3D and 4D seismic technology, the Schlumberger's interpretation software system and geophysical reservoir characterization and monitoring methodology .He has also coordinated several joint research projects between university - and industry - partners with national or European support . Lars has published more than 70 scientific papers and holds a number of patents.

Monday, December 10th, 2007

RULES FOR AN OPTIMAL FLOW SIMULATION GRID GEOMODELING AND FLOW SIMULATION: THE NECESSARY INTEGRATION

presented by Jean-Laurent Mallet
Emeritus Professor at Ecole Nationale Supérieure de Géologie
and
Emmanuel Gringarten
Product Manager at Paradigm

1) Rules for an optimal flow simulation grid – Jean-Laurent Mallet (speaker) and Emmanuel Gringarten

So far, in the oil and gas industry, stratigraphic grids have been used by two populations of geoscientists: the geostatisticians and the reservoir engineers. However, a close look on the respective needs of each of these geoscientists reveals that they do not have the same needs. For the geostatisticians, the (i,j,k) indexing of the nodes of the cells is used as a discretisation of an implicit curvilinear coordinate system. For the reservoir engineer, the grid is used to approximate numerically differential equations and the (i,j,k) indexing of the nodes is only used to retrieve the adjacent cells of each cell. It is clear that we do not need an (i,j,k) indexing for retrieving the neighboring cells of a cell: this is in particular the case when an unstructured grid is used where the (i,j,k) indexing is then not possible. In practice, building (i,j,k)-indexed grids is often not possible in the presence of a complex fault network and it is then necessary to simplify, dramatically, this fault network. In other words, due to a particular, and arguable, technique used by the geostatisticians to parameterize the subsurface, the entire numerical geosciences community has entered, since a decade, into the job of violating mother nature for the only purpose of building these terrible (i,j,k) grids. There are today two (r)evolutions which should change dramatically this insane current practice: on one hand, it is now possible for geostatisticians to parameterize the subsurface without using any (i,j,k) grid and, on the other hand, the new generation of flow simulators does not care about (i,k,j) indexing of cells. Based on these considerations, in this article we analyze the specifications of stratigraphic grids which are really needed by flow simulators to approximate, numerically, the flow equations.

2) Geomodeling and flow simulation: the necessary integration – Emmanuel Gringarten (speaker) and Jean-Laurent Mallet

The tasks of geomodeling and flow simulation have for far too long been dealt with isolation from one another – the argument being that they make each other’s processes too cumbersome. However, it is recognized that neither has validity without the other and at the very least a direct link must exist between the two.
The 3D modeling environment provides ways to integrate all sorts of geophysical and geological data to provide a numerical representation of the reservoir that is then used as input to the engineers’ flow simulation. Today, data exchange is currently accomplished through awkward file exchange formats that lose all knowledge relating the creation of the final 3D model. Nonetheless, this is the model that will be used by engineers to estimate reserves and optimize development plans.
There are many reasons why a seamless integration of geomodeling and flow simulation processes is needed. On the geosciences side, it is important to understand the dynamic impact of the geological features that make up the earth model, it is also necessary to ensure that the various geological scenarios considered yield a sufficient spread in flow responses for the purpose of production uncertainty forecasting. If only a few models can be carried forward to in-depth reservoir engineering analysis, these need to be screened and ranked with respect to their dynamic response. If upscaling is involved, the coarser reservoir model should exhibit the same flow characteristics as the finer-scale model. But above all, new field management processes relying on history matching, development optimization, and production uncertainty must consider the geological model and its associated uncertainty.
These themes and more will be developed during the presentation.

Jean-Laurent Mallet

Jean-Laurent Mallet graduated from the Ecole-Nationale-Supérieure-de-Géologie (Nancy/France) in 1968. He diversified next his interests into applied mathematics, finishing his research education with a PhD in applied mathematics and computer sciences at the Institut-National-Polytechnique-de-Lorraine (INPL). In his early years, he worked with the CNRS in the field of numerical data analysis and automated mapping. In 1981, he was nominated Professor at the Ecole-Nationale-Supérieure-de-Géologie, from which he directs teaching and research of applications of computer sciences related to geology.

Professor Mallet’s research is at the frontier of mathematics, geosciences and the computer sciences. He has contributed in many ways to the progress of sciences and technology in the field of Oil and Gas. He headed the Gocad Research Consortium since 1989 to 2006. He was awarded several international prizes such as the “Italgaz prize” (Italian Academy of Sciences), the “Antony-Lucas gold medal” (SPE) and the “grand prix Dolomieu” (French Academy of Sciences).

Professor Mallet is now “Emeritus Professor” at the Ecole-Nationale-Supérieure-de-Géologie and “Research Fellow” at Paradigm-Geophysical.

Emmanuel Gringarten

Emmanuel Gringarten is Product Manager at Paradigm for the Reservoir Product Line of the Gocad Suite, focusing on the development of workflow-driven integrated applications for consistent construction of reliable 3D reservoir models. He holds a BSc in Mathematics from Imperial College, London, and Masters and PhD in Reservoir Engineering from Stanford University. Emmanuel has been involved in many reservoir modeling projects, specializing in geostatistical and uncertainty modeling.

Tuesday, December 11th, 2007

GRIDDING

presented by Ivar Aavatsmark
Principal Researcher and VISTA Professor
Centre for Integrated Petroleum Research (CIPR)
University of Bergen
and
Zoltan Heinemann
Emeritus Professor
University of Leoben

Control-volume methods are prevailing for fully implicit solution of multiphase flows in reservoir simulation. Traditionally, reservoir simulators use  corner-point geometry in a logically Cartesian grid. There are two reasons for  this choice. Firstly, corner-point grids are well suited to describe the geological layering in reservoirs, and secondly, logically Cartesian grids yield a system of equations whose matrices have a band structure suitable for linear solvers. Since the linear solver is the most  time-consuming part in black-oil reservoir simulation, the last property is important.

However, more general grids are better suited to describe complex geological structures, such as regions with faults or fractures and regions with skew wells in layered media. In the last decade, some simulators  have been extended to handle such grid types. Unstructured grids  and radial-type grids have become popular in near-well regions.  Unstructured grids are also used in the neighborhood of complex geometric  structures, while the modeling of the fractures requires special techniques.

Computationally, it is advantageous if the grids can be generated such that they become K-orthogonal (orthogonal with respect to the permeability tensor K), but this may be difficult for complex geological structures or for skew wells in layered media.

To handle non-K-orthogonal grids, multipoint flux approximation (MPFA)  methods were introduced. In these methods, the flux through an interface  is approximated by taking the pressure difference between several  grid points (cell centers). This is an extension of the conventional two-point  flux approximation (TPFA) method, where the pressure difference between only  two points is used.

When TPFA methods are used for non-K-orthogonal grids, the solution looks nice  and it is likely to converge to the solution of a physical problem. However, it will not converge to the solution of the correct physical problem. On the other hand, when MPFA methods are applied for non-K-orthogonal grids, the solution  will converge to the solution of the correct physical problem (provided the  grid is not too extreme). However, although the solution converges to the  correct physical solution, it may contain unphysical oscillations. To avoid  such oscillations, the method should be monotone.

For TPFA methods, error estimates when using non-K-orthogonal parallelogram  grids are presented. These estimates indicate for which grids TPFA methods  may be used for. 

For MPFA methods, convergence rates for rough quadrilateral grids are given.  To avoid unphysical oscillations, only monotone methods should be used.  However, MPFA methods are only conditionally monotone. Monotonicity domains  are given for different variants of the method for quadrilateral grids,  indicating that multipoint methods which use as few points as possible to  express the flux are optimal.

Ivar Aavatsmark

Ivar Aavatsmark is a principal researcher and VISTA professor at the Centre for Integrated Petroleum Research (CIPR) at the University of Bergen. He holds a PhD degree in numerical mathematics from the Norwegian Institute of Technology, Trondheim. Aavatsmark worked twenty years at the research center of the oil company Norsk Hydro before he joined CIPR. He has published papers in discretization for elliptic and hyperbolic partial differential equations, well performance and compositional reservoir simulation. He has also written a monograph on thermodynamics of mixtures. His main interests lie in control-volume methods for general grids.

Zoltán E. Heinemann Professor Emeritus and former Head of Petroleum Engineering Department at the Mining University, Leoben, Austria. Holds a M.Sc. and a Ph.D. in Petroleum Engineering. He has 35 years experience in consulting and R&D in the Petroleum and Reservoir Engineering field with key qualifications in Reservoir Engineering, Reservoir Simulation and Enhanced Oil Recovery. He is the founder of HOT and creator of the SURE simulator. He has authored of more than 80 publications and several technical textbooks.

Zoltán E. Heinemann

Zoltán E. Heinemann Professor Emeritus and former Head of Petroleum Engineering Department at the Mining University, Leoben, Austria. Holds a M.Sc. and a Ph.D. in Petroleum Engineering. He has 35 years experience in consulting and R&D in the Petroleum and Reservoir Engineering field with key qualifications in Reservoir Engineering, Reservoir Simulation and Enhanced Oil Recovery. He is the founder of HOT and creator of the SURE simulator. He has authored of more than 80 publications and several technical textbooks.

Tuesday, December 11th, 2007

SOLVING RESERVOIR SIMULATION EQUATIONS

presented by Klaus Stueben
Fraunhofer Institute SCAI
Sankt Augustin, Germany

Reservoir Simulation is a necessary tool for making decisions. It is through numerical simulation that one can obtain knowledge pertaining to the processes occurring in the interior of an oil reservoir and hence, to enable an analysis of the various recovery strategies in order to guarantee optimal exploitation. History matching, optimization, and what-if scenarios require the simulation of large numbers of complex field-scale models.

Deep inside any reservoir simulator, large linear systems of equations need to be solved over and over again. During the last years, reservoir models have been growing in geometric complexity, heterogeneity, and size, causing these linear systems to get increasingly large and difficult to solve by classical numerical approaches. In fact, the computational time required to solve these systems of equations is today's major bottleneck in the practicability of numerical simulation. Hence, today's advanced reservoir simulators necessarily require particularly fast numerical solver modules to yield answers in an economical time frame.

The so-called algebraic multigrid (AMG) technique provides a state-of-the-art means to solve large applications with highest efficiency, usually much faster than any classical approach. As a consequence, AMG-based approaches are becoming increasingly popular as numerical kernel in many industrial simulation codes. AMG's major advantages - not only numerical efficiency but also robustness, scalability and ease-of-use - have become the driving forces behind its growing success in industrial use. A state-of-the-art software package, SAMG, developed at the Fraunhofer Institute SCAI, is currently being used in such diverse areas as fluid flow, structural mechanics, oil reservoir and ground water simulation, casting and molding, process and device simulation in solid state physics, electro-chemistry, and circuit simulation. In particular, in oil reservoir simulation, SAMG has become a well-established tool for various software providers as well as major oil companies.

A short introduction will be given to AMG, focusing on the key aspects that make it scalable and robust. We will present examples of large reservoir simulations enabled with AMG. Finally, research activities to continually improve and expand AMG's capabilities will be outlined.

 

Klaus Stüben

Klaus Stüben is Deputy Head of the Numerical Software Department of Fraunhofer's Scientific Computing Institute, and is responsible for the scientific development of modern fast numerical solvers. He is one of the inventors of the algebraic multigrid (AMG) technique and is heading the development of the advanced software library, SAMG, which has become a well-established tool in many industrial simulation systems. Dr Stüben is currently focussing on the development of efficient numerical approaches for solving various types of partial differential equations including such diverse areas as fluid flow, structural mechanics, oil reservoir and ground water simulation, casting and molding, process and device simulation in solid state physics, electro-chemistry, and circuit simulation. Dr Stüben has obtained his masters and doctors degrees in Mathematics at the University of Cologne.

Wednesday, December 12th, 2007

Modeling wells and production facilities

presented by Jonathan Holmes
Scienti.c Advisor at Schlumberger

The primary purpose of a reservoir simulator's well model is to supply source and sink terms to the reservoir model. It must determine the flow rates from each of the reservoir grid cells that contain the well completions while the well is operating under a variety of different constraints. It must also determine the composition of the wellbore contents, so that if crossflow occurs the correct mixture can be reinjected back into the reservoir. It will be shown how a conventional well can be modeled as a set of mass balance equations for each component, together with an equation describing the operating constraint on the well.

The use of advanced wells (horizontal, multilateral, and smart wells containing flow control devices) has imposed additional demands on reservoir simulation well models. They must allow the fluid mixture properties to vary with position in the well, and enable different fluid streams to comingle at branch junctions. Friction may make an important contribution to the local pressure gradient. To provide an improved representation of the physics of fluid flow, we subdivide the well into a network of segments, where each segment has its own set of variables describing the local fluid conditions (pressure, flow rate, composition). Individual segments can be configured to represent flow control devices, accessing lookup tables or built-in correlations to determine the pressure drop across the device as a function of the flow conditions.

The ability to couple the wells to a production facility model such as a pipeline network is a crucial advantage for field development and optimization studies, particularly for offshore fields. The existing techniques for coupling a facility model to the reservoir simulation model fall into two categories. In one category the facility model and the well/reservoir model are integrated into the same simulation code. The complete system of equations may then be solved simultaneously. In the other category the models are run as separate simulations that are coupled by the exchange of boundary conditions. This allows the engineer to use a choice of specialist facility simulators. We discuss the issues concerning each of these techniques.

Jonathan Holmes

Jonathan Holmes is a Scientific Advisor at Schlumberger. His principal area of interest is reservoir simulation, in particular the modeling of wells and surface facilities, multiphase fluid flow and field production/injection strategies. He has a long association with the ECLIPSE reservoir simulator, from its inception at ECL through its subsequent development within Intera and Schlumberger. He is now on a project team developing a new reservoir simulator. Before entering the petroleum industry he worked on fluid flow simulation in nuclear reactors for the UK Atomic Energy Authority. There he went on to work with Dr Ian Cheshire on the PORES reservoir simulator before moving with him to ECL. He holds a BA degree in physics and a PhD Phil. in astrophysics from Oxford University.

Wednesday, December 12th, 2007

MODERN TECHNIQUES FOR HISTORY MATCHING

presented by Ralf Schulze-Riegert
principal consultant with
Scandpower Petroleum Technology
and
Dr. Shawket G. Ghedan
Associate Professor of PE at the Petroleum Institute of Abu Dhabi and a Schlumberger-NExT Instructor

A number of structured workflows for uncertainty assessment have been presented over the past years. Most documented procedures focus on reservoir field evaluations with no or few constraining dynamical measurement data, i.e. production data or 4D seismic. Established workflows exist for static modelling including multiple scenario evaluations and statistical procedures for uncertainty quantification in terms of cumulative distribution curves.
It is generally accepted that any model realistically predicting unknown future quantities should reproduce known history data. This requires a model validation process called History Matching, which is traditionally cumbersome and time consuming. A consistent introduction of production data and/or 4D Seismic is computation-intensive. Against this background, new approaches in the application of various experimental design and optimisation methods are called for which are supported by the use of high-performance computing facilities.
This talk provides an overview of “Modern Techniques for History Matching”. Workflows are compared and preferences for particular application scenarios are explained.
The following aspects will be covered in detail:

• Recent development efforts and integration of optimisation techniques into the reservoir engineering workflow

• Efficient use of Optimisation and Experimental Design Techniques in a case-specific environment

• Event-Targeting-Model-Calibration to handle large and complex simulation cases

• History Matching as an input to Risk Assessment

• Scenarios and framework for structured workflow processes

Ralf Schulze-Riegert

Ralf Schulze-Riegert is a principal consultant with Scandpower Petroleum Technology. With a background in theoretical physics, Ralf Schulze-Riegert holds a PhD degree from the University of Hanover, Germany. For more than ten years he has worked as a consultant in the Oil & Gas and Automotive industries with a focus on project management and software development.
Ralf Schulze-Riegert has a special interest in mathematical modelling and optimisation techniques. Since 2005 he has been a part-time lecturer at the Technical University of Clausthal, Germany, specialising in “Model Validation, Computer-Aided Design and Optimisation”.
At SPT he is currently Product Development Manager of the MEPO optimisation tool and is responsible for developing and evaluating optimisation techniques and new workflow scenarios. With application to reservoir simulation, he regularly presents and co-authors research findings and case studies on model validation, optimisation and uncertainty quantification.

Dr. Shawket G. Ghedan

Dr. Shawket G. Ghedan is an associate professor of PE at the Petroleum Institute of Abu Dhabi and a Schlumberger-NExT Instructor. He has over 20 years of diverse academic and industrial experience. Before joining the PI, he worked with different oil and gas operating companies, dealing with reservoir operations and management, reservoir characterization; static and dynamic modeling as well as long term development planning studies of some major carbonate oil fields. He is the principal author of a number of technical papers. Research areas include: modeling of oil recovery in the transition zones, dynamic rock typing, Modeling of triple porosity fractured reservoirs and stochastic optimization techniques for HM and full filed development. Dr. Ghedan is the chairman of Abu Dhabi SPE Section. He was the 2005-2006 SPE distinguished lecturer. He holds a B.Sc. degree in Petroleum and Minerals Engineering from Baghdad University, M.E. and Ph.D. degrees in Reservoir Engineering from the Colorado School of Mines.

Wednesday, December 12th, 2007

STREAMLINE BASED HISTORY MATCHING AND PREDICTION

presented by Marco Thiele
President of Streamsim Technologies, Inc.

Streamline-based flow simulation (SL) is an effective and complementary technology to more traditional flow modeling approaches such as finite differences (FD). This is because streamline-based flow simulation is particularly efficient in solving large, geologically complex and heterogeneous systems, where fluid flow is dictated by well positions and rates, rock properties (permeability, porosity, and fault distributions), fluid mobility (phase relative permeabilities and viscosities), and gravity. These are the class of problems more traditional modeling techniques have difficulties with. More diffusive mechanisms, such as capillary pressure and expansion-dominated production, on the other hand, are not modeled efficiently by streamlines.

Modern SL simulation rests on 6 key principles: (1) tracing three-dimensional (3D) streamlines in terms of time-of-flight (TOF); (2) recasting the mass conservation equations along streamlines in terms of TOF; (3) periodic updating of streamlines; (4) numerical 1D transport solutions along streamlines; (5) accounting for gravity effects using operator splitting; and (6) extension to compressible flow. These principles are reviewed here.

The application of SL simulation is presented in the context of what are generally considered central issues in reservoir engineering: (1) (water) flood optimization; (2) history matching; (3) quantifying uncertainty in reservoir performance; (4) upscaling; (5) computational speed; and (6) complex flow mechanisms such as miscible flooding and dual porosity systems. Particular attention is paid to novel, streamline-specific data that add valuable engineering insight, as in the case of injector/producer efficiencies and as an aid in upscaling.

The speed and efficiency as well as the availability of new data make streamlines potentially the most significant tool for solving complex optimization problems related to history matching and optimal well placements. Finally, the outlook for streamline-based simulation is discussed in the context of the next generation of simulation tools, geologically consistent history matching, dual-media systems, and parallel computation.

Marco Thiele

Marco Thiele is president of Streamsim Technologies, Inc., a company he co-founded with Dr. Rod Batycky and Prof. Martin Blunt in 1997 to commercialize reservoir simulation technology based on streamlines. Starting in 2006, Dr. Thiele is also a Consulting Professor in the Department of Energy Resource Engineering at Stanford University. Previously, he was an Acting Assistant Professor at Stanford University teaching graduate-level courses on Reservoir Simulation, Thermodynamics of Phase Behavior, and Applied Mathematics in Reservoir Engineering. His research at Stanford focused on streamline-based flow simulation, uncertainly in reservoir forecasting, and integrated reservoir management. Dr. Thiele consults widely on reservoir simulation in general, and streamline simulation in particular, and how SL-based flow simulation can help companies improve the quality of reservoir studies while reducing project turn-around times.

Dr. Thiele has authored a number of papers on streamline simulation and is a frequently invited lecturer and keynote speaker on the subject. Dr. Thiele is also a technical editor for the SPE Reservoir Evaluation and Engineering Journal and serves on the SPE Primer Book Committee.

Dr. Thiele received his PhD in Petroleum Engineering from Stanford University in 1994 and his Masters and Bachelors from the University of Texas at Austin in 1989 and 1986 respectively. In 2004, Dr. Thiele completed his Business Administration Certificate from the University of California, Berkeley Extensions. He is the recipient of the 1996 SPE Cedric K. Ferguson Medal, winner of 1994 International SPE Student Paper Contest, and a 1991 distinguished SPE speaker invited by the SPE Adriatic Section.

Thursday, December 13th, 2007

OPTIMIZATION OF FIELD DEVELOPMENT

presented by Burak Yeten
Reservoir Simulation Engineer
of Energy Technology Company of Chevron

To maximize reservoir performance (recovery or net present value), a series of algorithms are developed to optimize the number of producers and injectors, their types (e.g., vertical, horizontal or multilateral), locations, trajectories, scheduling and operating principles. The effects of uncertain geological and engineering parameters are included in this optimization.

The determination of the optimal type, location and trajectory of a nonconventional well is very challenging. The problem is more complicated than other well optimization problems because of the wide variety of possible well types (i.e., number, location and orientation of laterals) that must be considered. The optimization procedure entails a Genetic Algorithm applied in conjunction with several acceleration routines that include an artificial neural network, a hill climber, and a near-well upscaling technique. The overall methodology is then applied to a number of problems involving different reservoir types and fluid systems.

The scheduling of the wells that might be predrilled is also considered. In this case the problem is treated as a Traveling Salesman Problem (TSP) using a genetic algorithm. This algorithm was accompanied by various proxy and helper functions to speed up the convergence and reduce the CPU time. This algorithm is applied to various cases. The cases studied showed benefits due this optimization.

The operating principles of the wells are also optimized using the aforementioned algorithms. These principles include production/injection pressures/rates, voidage ratio, workover operations, etc.

The different optimization problems described above are all solved independently. Depending on the requirement of a particular problem these different optimizations can be combined and run concurrently to optimize a field performance. Examples of these combinations are also presented.

Burak Yeten

Burak Yeten holds BS. and MSc. degrees from Middle East Technical University and a PhD from Stanford University, all in petroleum engineering.

Burak Yeten currently works as a senior reservoir simulation engineer at the Resevoir Simulation Consulting Team of Energy Technology Company of Chevron, based in San Ramon, California, USA. Prior to joining Chevron he worked as a reservoir engineer at the International Projects Group of Turkish Petroleum Company. His research interests include optimization of development and management of petroleum reservoirs, history matching, uncertainty assessment, decision analysis and deployment and control of smart wells.

He is a member of the Society of Petroleum Engineers. He served as a technical editor with Society of Petroleum Engineers Reservoir Engineering and Evaluation Journal for more than two years and is currently serving to the same journal as a review chairperson.

Thursday, December 13th, 2007

CLOSED-LOOP RESERVOIR MANAGEMENT

presented by Jan Dirk Jansen
Delft University of Technology and Shell International
Exploration and Production

The lecture will address concepts and computational tools to enable ‘closed-loop’ model-based reservoir management. Also known as ‘real-time’ reservoir management this involves the use of (uncertain) reservoir and production system models in combination with production measurements and data from other sources such as time-lapse seismics to continuously update the models. The key sources of inspiration for our work are measurement and control theory as used in the process industry and data assimilation techniques as used in meteorology or oceanography. Important steps towards full closed-loop reservoir management are the development of optimization, data assimilation and model reduction techniques, and in particular their integrated application in a reservoir management workflow. Methods to be discussed will include optimal control theory (adjoint-based optimization) and the Ensemble Kalman Filter. Numerical examples will be presented to illustrate the scope for model-based optimal control of waterflooding using real-time production data under uncertain reservoir conditions.

Jan Dirk Jansen

Jan Dirk Jansen has a joint assignment (50%-50%) at the Delft University of Technology and Shell International Exploration and Production. In Delft he holds the Chair in Reservoir Systems and Control at the Department of Geotechnology, while at Shell he is a Consultant in the Exploratory Research group. His research interests focus on the use of systems and control theory as applied to well bore and reservoir flow. He holds MSc and PhD degrees from Delft University of Technology, and has spent 21 years with Shell in research and operational positions in the Netherlands, Norway and Nigeria.