SPE logo
Registration travel and accommodations


Session 1: HPC Trends and Fast Solution Methods (Linear and Nonlinear)

Session Chairs: David Keyes, KAUST; Kees Vuik, Delft University of Technology

A primary challenge for reservoir simulation is the accurate description of multiphase flow in porous, highly heterogeneous and fractured media with very complex geometries. The lack of robust and efficient solvers for fully implicit formulations is still one of the main bottlenecks that most simulator developers face in the oil and gas industry. Generally, the underlying systems of partial differential equations are nonlinear, highly non-symmetric and indefinite. The condition number and degree of coupling of these systems may be subject to dramatic changes due to abrupt flow variations induced by the high-heterogeneity and complex well operations during the simulation process.

In addition, over recent years, reservoir models have been growing in complexity (regarding geometry, discretisation and physical models including, for instance, thermal and chemical influences, geomechanics), heterogeneity and size, causing these systems to get increasingly large and difficult to solve. In fact, the computational time required to solve these systems of equations is today's major bottleneck in the practicability of numerical simulation. Hence, advanced reservoir simulators need to combine the numerical scalability (optimality) of efficient solvers with the parallel scalability of modern hardware.

This session addresses the efficient use of parallel architectures (including both, large scale clusters of conventional x86 multi-core CPUs as well as heterogeneous systems equipped with hardware accelerators, e.g. GPUs), and all relevant numerical aspects such as the treatment of linear and nonlinear problems, treatment of fully coupled systems (either the Jacobian directly or by CPR-type approaches), efficient preconditioning techniques and fast linear solvers (hierarchical solvers such as multigrid or multilevel ILU).


Efficient and Robust Multilevel Simulation in the Presence of Heterogeneity and Uncertainty

Robert Scheichl – University of Bath


Advanced AMG Application for Solving CPR-type Pressure Systems

Sebastian Gries – Fraunhofer SCAI


A Class of Multilevel Preconditioners for Petroleum Reservoir Simulation

Jinchao Xu – Pennsylvania State University


Coffee Break      


MPI/Thread Parallelization of the CPR Preconditioner for Multicore Architecture

Pascal Henon – TOTAL


Multicore and Multi-GPU Parallelization of a General Purpose Reservoir Simulator

Yifan Zhou – Stanford University


Fast and Robust Solvers for Pressure Systems on the GPU

Kees Vuik - Delft University of Technology


Session 2: Multiphysics and Complex Model Challenges

Session Chairs: Bret Beckner, ExxonMobil; Michel Kern, INRIA

This session seeks to present challenges and solutions to problems associated with reacting and non-traditional fluid flows in porous media with particular emphasis on application in computational large models. Topics of interest include:

  • Challenges and solution methods for modelling reactive transport in porous media. Examples include models for porous media dissolution and/or deposition, induced wettability alteration, salinity effects on EOR and water flooding.
  • Modelling of localised, small scale phenomena within reservoir simulation grid blocks. Many important processes occur over length and time scales below that of a typical reservoir simulation. Modelling of combustion and solvent fronts, coupled wellbore flow and near well-bore flows such as fine migration and hydraulic fractures are examples.
  • Integrated and efficient modelling of coupled geomechanics and porous media flows
  • Advanced methods in compositional reservoir simulation including aggregation methods for large multi-field, multi-fluid systems, fast and robust algorithms for multi-phase, non-isothermal flash calculations and new methods for phase equilibrium calculations, for hydrocarbon/water/CO2 systems.
  • Modelling flows in large unconventional gas and oil deposits including modelling multi-phase dispersion and diffusion effects as well as models for CO2 sequestration.
  • The simulation of coupled models presents specific challenges in addition to those already found in the individual models. The session will also discuss coupling algorithms, their robustness, as well as how they can be adapted for high performance computing.
  • Field examples of the need for these complex, multi-physics models in large scale reservoir simulations will also be discussed.


Domain Decomposition Methods in Geomechanics

Adolfo Rodriguez - ConocoPhillips


Accelerated Dynamics Modeling o fthe In-Situ Upgrading Process

Jeroen Vink – Shell


Numerical Modeling of Large Scale Fluid Injection in Fractured and Heterogeneous Reservoirs

Abbas Firoozabadi – Yale University


Coffee Break      


Panel Session A: Spatial and Temporal Discretisation for Coupled Non-Linear PDEs

Moderators: Mary Wheeler, University of Texas at Austin; Serdar Celebi, Istanbul Technical University


Brett Beckner, ExxonMobil

Larry Fung, Saudi Aramco

David Keyes, KAUST

Hector Klie, ConocoPhillips


Session 3: Multiscale Modelling / Applications

Session Chairs: Ruben Juanes, MIT; Ivan Yotov, University of Pittsburgh

Hydrocarbon recovery processes act at many scales in a reservoir including the pore/capillary scale (for single-, dual-, or triple-porosity systems), the geologic laminae scale, the core scale, the geologic bed scale, the imposed numerical grid scale and the reservoir scale. The storage and transport mechanisms associated with these hydrocarbon recovery processes may be dominated by different physical phenomena at each of these scales. The objective of this session is to discuss methods to model hydrocarbon systems at multiple scales. Topics to be considered include numerical multiscale techniques, characterisation of reservoirs at multiple scales for model input, unconventional (shale gas/oil) reservoirs, upscaling, and uncertainty analysis at multiple scales.


Comparison of Unstructured Grids and Structured Grids for Flow Dynamics in Reservoir Simulation
Larry Fung – Saudi Aramco


Mimetic Finite Differences for Complex Geometries with Multiphysics

Omar Al Hinai - University of Texas at Austin


Developing Multiscale Finite Volume Algorithms for a "Next Generation" Reservoir Simulator

Seong Lee – Chevron


Coffee Break      


Reducing Numerical Dispersion in Reservoir Simulations: Moving Explicit Discretization

Bilgin Altundas – Schlumberger


Scalability of SuperLU Solvers for Large Scale Complex Reservoir Simulations

Sedar Celebi and Ahmet Duran – Istanbul Technical University


Mathematical and Computational Modeling of Coupled Surface

Ivan Yotov – University of Pittsburgh


Session 4: Large Scale Data Reduction/Compression, Mining and Visualisation

Session Chairs: Klaus Stuben, Fraunhofer Institute SCAI; Knut Andreas Lie, SINTEF

Data generated by observation and numerical simulation are the key to understand, explore and optimise oil and gas reservoirs. Results of many simulation runs are stored, archived, analysed and compared among each other. This puts very demanding requirements on storage systems, networks and visualisation systems. Approaches to overcome these issues are application specific data compression of single or multiple simulation results, data reduction methods and sensitivity analysis. These technologies are enabled to support data mining of simulation results, for example, to find all simulation results with a similar behaviour of a reservoir. This session aims at presenting technologies for data compression of CFD applications as well as data reduction methods, which allow identifying major trends in sets of simulation results and allow for (non-linear) interpolation replacing computation with a new set of parameters.

Human interaction with complex systems, e.g. for seismic interpretation or for reservoir modelling, relies on a good visualisation, helping users to explore and understand data, and also communicate that understanding to others. Visualisation is a central tool in carrying out analysis, enabling researchers and other users to explore datasets to identify patterns, associations, trends and more. A particular challenge again in terms of data volume and decision support is the analysis of several simulation results representing different loads or uncertain geologic formations. A good data visualisation of sets of results can help users to make robust decisions based on the data being presented. It should provide an effective representation of the underlying data, to help answer a particular question at hand.
Communicating data in this way can support senior decision-makers engaged in strategic planning, service managers needing to understand where delivery could be improved and managers wanting to monitor performance. This session shall provide answers to related challenges in major areas: Information display and decision aiding, progress in human-system interaction, data reduction and data compression.


Dr Lim Goh – SGI


Data Compression for Simulation Results

Clemens-August Thole – Fraunhofer SCAI


Improving Large Scale Visualization of Reservoir Models

Stein Dale – Ceetron


Analysis of Simulation Data with Nonlinear Machine Learning Methods

Jochen Garcke – University of Bonn


Coffee Break      


Panel Session B: Challenges in Reservoir Simulation

Moderators: Khalid Aziz, Stanford University; Philippe Quandalle, IFP Energies Nouvelles


Ali Dogru, Saudi Aramco

David Keyes, KAUST

Garfield Bowen, Schlumberger

Gregory King, Chevron

Jeroen Vink, Shell

Shuhong Wu, PetroChina


Session 5: Reservoir Management, Optimisation, Inverse Problems and Uncertainty Quantifications

Session Chairs: Hector Klie, ConocoPhillips; Gregory King, Chevron

This session will deal with the problem of estimating the distributions of porosity, permeability and geological shapes (like reservoir boundary position) in heterogeneous and multiphase oil and gas reservoirs, by matching their dynamic behaviour. The dynamic data is in the form of field measurements from well testing, production history, interpreted 4-D seismic information. This process, called history matching by reservoir engineers, is an inverse or parameter estimation problem: inferring reservoir parameters from the indirect measurements. During the session we will discuss on parameterisation techniques, objective function and inverse problem formulation for devising effective optimisation strategies. Strategies avoiding expansive gradient evaluations and solving adjoint problems will be in the scope of this session.

Finding the optimal location of wells may increase significantly a field development project Net Present Value (NPV). This problem is nowadays one of the most challenging ones in oil and gas field development. Because of the very large number of wells involved in the production of giant reservoirs, global optimisation may be a precious help to select the best well patterns, well spacing in patterns, order of drilling different zones. The session will also address global optimisation techniques and computational effort necessary for their implementation (in terms of number of reservoir simulations and associated computing time).

Model parameters being highly uncertain in the field of geosciences, managing uncertainties is also a major stake in the process of decision-making at every stage of reservoir production from field exploration to production. We will also discuss the development of statistical methods allowing for the evaluation of uncertainties impact and realisation of sensibility analysis on model parameters, with the aim of weighting uncertainties on reservoir production forecast.


The Source Method Application in Full Scale

Vitaly Baikov – Rosneft


ES-MDA as a Replacement for EnKF for History Matching and Uncertainty Quantification Workflows

Al Reynolds – University of Tulsa


A Stochastic Ensemble Method for Parmeter Estimation

Ahmed El Sheikh – University of Texas


Robust Optimisation of CO2 Sequestration Strategies Under Geological Uncertainty

Rachares Petvipusit – Imperial College London


Coffee Break      


An Efficient Workflow Using Mixed Genetic Algorithm

Masroor Chaudhri – Chevron


An Adaptive Hierarchical Metamodelling Approach for History Matching of Reservoir Simulation

Tanja Clees – Fraunhofer SCAI


Optimal Rate Control Under Geologic Uncertainty

Akhil Datta-Gupta – Texas A&M


Session 6: Lessons in Large Scale Computing in E&P and Other Application Areas

Session Chairs: Patrick Demichel, Hewlett Packard; Ulisses Mello, IBM

As we progress towards Exascale computing at the end of the decade, we can observe that numerically intensive large-scale applications in several domains of science and business face numerous challenges to achieve three orders of magnitude performance improvement within 10 years. This will enable unprecedented application capabilities for a 21st century and will open up new frontiers of fundamental research spanning exploration and production, energy, life sciences, engineering, climate, environment, finance and materials science. To achieve this ambitious goal, massive improvement in current state-of-the-art of power efficiency, scalability and reliability of computing systems is required, along with exploitation of new technologies like non-volatily memory, silicon photonics and three-dimensional packaging. Lots of progress has been achieved, unfortunately Exascale will be radically different in many aspects to what we have learnt in the race to Petascale, in particular because we have reached the economical, power, complexity limits of all data centers/institutions. Some emerging problems could become rapidly critical inhibitors, like the data explosion. It is important to understand how new technologies will affect our architectures and how to adapt current and new emerging applications to obtain the greatest benefit of the potential at the lowest effort of transformation and minimise the pressure on the developers. For the designers it is very critical to understand how the code loads and data sets will evolve in the next generation of problems we expect to solve. Then a co-design effort hard/soft is a fundamental dimension of the success of the Exascale effort. In this session, we will discuss technologies and algorithmic transformations to make some fundamental breakthroughs that are necessary to reach the next frontiers.


Keynote Address: Reservoir Simulation of Santos Basin Giant Pre-Salt Fields

Jorge Pizarro – Petrobras


A Massively Parallel, Scalable, Simulator for Giant Reservoirs

Usuf Middya, Saudi Aramco


History Matching of Giant Reservoir Models

Kirill Bogachev – Moscow State University


Coffee Break  


Panel Session C: Exploitation of Nonhomogeneous Processing: Extreme Computing Hardware and Software Roadmap, and the Road to Exaflop Computing

Moderators:Philippe Thierry, Intel; Motoi Okuda, Fujitsu


Hybrid Computing Architecture for Multi-billion Cell Reservoir Simulation

Gunter Roeth, Bull


Exascale Computing Research: Focus on a Tool Chain Performance Analysis

Jean-Thomas Acquaviva - Exascale Computing Research Center


Exascale Vision with HP Labs

Patrick Demichel – HP