A bold advance in reservoir simulation performance

advertisement
A bold advance in reservoir
simulation performance
Parallel Interactive Reservoir Simulator
The industry first integrated modeling and data analysis framework centered
on dynamic reservoir simulations combining interactive 3D graphics and
supercomputer parallel performance with no extra per core charge.
Parallel Interactive Reservoir Simulations
Major Reasons to Apply tNavigator®
Universal software: from sector waterflood and
history matching to running giant models on cluster
Black-oil
Compositional
Thermal Compositional
Computational performance
• Turns your workstation into an interactive desktop
supercomputer with 10–20 times parallel performance boost.
• The license price is per computer and does not
depend on the number of CPUs, cores used.
• Clusters: Revolutionary Hybrid technology.
• Record breaking parallel scalability: from x 100 to
x 1300+ acceleration factor.
• Clouds: scalable ‘pay-per-use’ solutions.
All basic simulation model features including
• Multi-component, fully-implicit calculations, dual
porosity, dual permeability.
• Corner point, general mesh formulation (NNC, LGR,
faults, pinchouts, etc.), unstructured grids.
• THP, drawdown well controls; unlimited hierarchic
group structure and controls.
• Multiple flow, PVT (including multibranching), equilibrium regions.
• Vertical, deviated, horizontal, multi-segmented
wells, fractures.
• Aquifers (Fetkovich, Carter Tracy, constant flux,
constant head pressure, numerical).
• Simulation of tracers.
• Numerically efficient EOR modeling toolbox: alkalines, surfactants, polymers.
• Surface network option, compressors, chokes, VFP
lifting tables.
• Flux options for sector modeling.
• Full field thermal simulations: steam injections,
SAGD.
New technologies
• Fractured well model.
• Quantitative estimate of waterflood efficiency.
• Fully-automatic split/merge models.
• Smart loading routines for “sparse” models.
• Wise approach to dual porosity/permeability simulation speed-up.
• Desalination for collectors with salt.
• Interactive tracers with activation conditions for
injection of chemicals.
Fully Interactive Simulations
• Monitor all 3D dynamic reservoir & well data at run­
time.
• Add wells, patterns, restart runs, generate forecasts
interactively at runtime.
• Pre-, post-processing, monitoring through one
intuitive multiplatform GUI.
• Runtime analytics, 3D streamlines, waterflood optimization.
Uncertainty quantification and assisted history matching
• Part of the simulation package. Easy workflow set
up.
• Experimental design methods.
• Optimization algorithms.
• Analytic toolbox.
Seamless support of the existing
industry standard input formats
• Converts E100, E300, Tempest MORE, IMEX, STARS
input deck formats “on the fly”.
• The simulation results can be saved as E100 binaries
to link the simulator with the software in your workflow.
Basic geological modeling features
• Handling of well log data.
• Arithmetics on the grid.
• Reservoir properties interpolation in the interwell
space.
• Deterministic and stochastic approaches.
• Kriging. Sequential Gaussian simulation method.
• Voronoi grids.
• Flexible well data format.
Comprehensively benchmarked against industry
standard packages.
Corporately used by international oil and gas
companies.
Maximize your reservoir
simulation performance
for free!
tNavigator is the industry leading parallel
dynamic reservoir simulator that efficiently
utilizes all available computational resources.
Based on the recent hardware developments, tNavigator provides record boost to reservoir simulation
performance. Smart balancing of the computational
resources of multi-CPU multicore systems allowed
breaking through the existing industry standard acceleration limits, and reaching almost an order of the
magnitude improvement in cluster performance.
• All elements of the simulator are parallel.
• System thread parallelization for multicore workstations.
• MPI algorithms for clusters.
• Superscalable hybrid technology for multicore CPU
clusters.
• Efficient memory utilization.
• Uniform load distribution between the cores.
• Non-uniform memory access.
No additional fee for multicore computations.
tNavigator software uses power of all the available CPU cores of the workstation.
We do not tax your hardware!
Interactive Reservoir
Simulation Tools
tNavigator offers reservoir engineers and geologists unique runtime monitoring of the simulation
runs, advanced field development planning and waterflood optimization functionality interactively with a
3D graphical user interface:
• Run dynamic reservoir simulations 10–20 times
faster with modern multicore computers.
• No painful data migration, since tNavigator directly
loads and exports industry standard ASCII and
binary formats.
• Monitor every dynamic grid property and well data
live at runtime, while running simulations on workstations or clusters.
• Easily handles small sector and large full-field models with minimum or no upscaling.
• Run full finite difference simulations, perform 3D
streamline-based waterflood analysis in the same
interface with no additional software to buy.
• Create and run forecast models instantly from any
time step, right after the history matching is completed.
• Load your well log data directly from LAS files and
plot it against the well production profiles
• Update your grid parameters based on 3D interpolation or kriging directly in the simulator without
going back to the geological package
• Novel hydraulic fracture simulations for full-field
models applied for generic, shale gas, and oil reservoirs.
• Efficient sidetrack optimization modeling.
• Split and merge dynamic reservoir model grids with
no additional geological packages needed.
• Visualize dynamic well production and injection
profiles perforation by perforation, at runtime.
• Switch wells from production to injection, add new
wells and well patterns, plan new horizontal well
trajectories, inject tracers with a few mouse clicks.
• Include simulations of near well bore treatments
into full-field modeling.
Parallel Interactive Reservoir Simulations
tNavigator ® provides fantastic scalability on shared
memory workstations and high-performance clusters
CPU computational efficiency grows rapidly today due to an increase in the number of cores. At the same time,
the cost of high-performance computational systems is constantly decreasing. Taking into account the availability
of computational power, there is a growing demand for efficient software, which would be able to utilize all the
computational resources in hand with a wise approach to parallel reservoir simulations. We know how to make it
efficient.
Shared Memory Thread Based
Parallelization
Direct thread based parallel algorithms are the most
efficient approach for workstations with multicore
CPUs. With minimal computation costs for data exchange between the cores, this technology takes full
advantage of CPU evolution.
Software & Hardware hand-in-hand: fast CPU cache,
Non-Uniform Memory Access (NUMA), Hyper-threading,
all parts of tNavigator are parallel (not just linear solver), special compiler settings, and many other tricks to
maximize the computational efficiency.
tNavigator Hybrid uses a conceptually new approach to parallel computations on clusters (SPE
162090, 163090). The main idea of the hybrid algorithm is to adopt different parallel algorithms for every
stage.
Parallel acceleration versus number of cores used for different
generations of CPU’s (including hyperthreading)
From days to hours!
Hybrid Technologies
for Multicore CPU Clusters
Over the last few years, the cluster system architecture has changed dramatically. These changes represent the most significant advance over the last 20
years. Every computational node of a cluster is now a
multicore computer with non-uniform access to RAM.
Losing sight of these changes during software development leads to a dramatic loss of efficiency and poor
CPU utilization.
• Cluster level: model is loaded into cluster and its
grid is divided between the cluster nodes according
to an equal number of active grid blocks.
• Node level: the parallelization between CPU cores
is done on the level of matrix.
• Simulator solver software integrates both MPI and
thread system calls.
• The number of MPI processes is limited to the
number of cluster nodes, not the number of cores.
The approach in tNavigator Hybrid
removes the bottlenecks in parallel
scalability!
Hybrid Performance Results
Giant Field Challenge
The algorithm performance was studied on the RFD
cluster with 20 nodes, 40 Xeon 5650 CPUs, 240 cores, 480
Gb DDR3 1333MHz RAM, QDR 4x Infiniband (40 Gb/s).
This is a portable cluster which can be placed in a small
server room.
The three phase black-oil model for one of the
world’s largest oil fields was used for tNavigator Hybrid
performance tests. It contains 43 million active blocks,
nearly 14 000 wells, and covers 42 years of production
history.
The huge number of well perforations used in this
reservoir model presents a very big challenge for parallel simulations. But even in this extreme environment, the hybrid algorithm shows excellent scaling
results, and exhibits continuous growth of efficiency.
processors, which sums to 4096 simulation cores.
A real field three-phase model with 39 wells and 10
years of history was chosen as the study test object.
The model contained 21.8 million active grid blocks.
The model calculation time was reduced from
2.5 weeks down to 19 minutes. The resulting speedup coefficient compared to one calculation core is
equal to 1 328! This can be called the world record for
commercial solutions in this field!
From days to minutes!
Cloud Simulations
The total simulation time for this model was
39 minutes compared to 61 hours with one CPU core
demonstrating a parallel acceleration factor of 94! The
acceleration factors for the less complex dynamic reservoir models on that cluster are found in the range of
80–120.
tNavigator Hybrid algorithm has been successfully
applied to a number of giant models with up to 400
million active grid blocks and 1.4 million well connections (SPE 171225).
Extreme Scalability Challenge
We have not seen algorithm saturation in any of the
calculation series. To check the limitations we applied
a more powerful 512 node cluster. Each cluster node
consisted of two four-core Intel® Xeon 5570 Nehalem
If you need to run more model realizations or get
the results sooner, you simply need to add more CPUs.
One way is to get new servers installed locally in your
company.
Another alternative is to rent the computational
power available in clouds. RFD has recently implemented a cloud simulation system available for all companies running the simulations. All it takes to run the
model from the user side is to upload the input data via
encrypted channels (similar to the banking online systems), chose the number of CPU’s to be rented and press
run. The simulation results can be observed remotely at
runtime from the user terminal located anywhere in the
company. The encryption key can only be used in the
company network, so the data is 100% secured when
simulating in remote clusters within the cloud.
This makes the reservoir modeling solutions much
more scalable in terms of time, money, human and computational resource.
Parallel Interactive Reservoir Simulations
Runtime 3D Visualization
Package
Simulation in progress!
• Preprocessing, run monitoring, and the simulation result post processing are done using a single
multi-window graphical interface working under
Windows and Linux.
• Start, pause, restart simulation runs and configure
multiple forecast models with a few mouse clicks.
• Every static and dynamic reservoir property in the
model can be viewed as 2D, 3D maps, histograms,
and user defined vertical profiles.
• Multifunctional 2D bubble maps to display well
production and injection instant and accumulated
rates, historical data and mismatches.
• Runtime graphs of rates and property changes with
time for each model grid block, for each well perforation.
• Dynamic contour lines of any 2D property map.
• Dynamic well production and injection profiles can
be visualized at the perforation level at runtime.
• Waterflood optimization tools: 3D and 2D streamlines are calculated on each time step from the
current pressure distribution, drainage matrix and
injection efficiency graphs, user arithmetic, etc.
• Flow maps on the grid: accumulated flux in the
gridblocks, influx through the faces at each time
step.
• Interactive tracer simulations.
• Advanced comparison tables and graphs for efficient history matching.
• Advanced user map calculator.
• Well log data comparison with production profiles
along the wellbore
• Well log correlation maps.
• Static model update in the GUI without going back
to geological software.
• Click of the button comparison of multiple simulation results.
The package allows interactive adjusting of the dynamic model in the GUI with immediate visualization
on maps and plots. All visualization features are available during the run for results monitoring. tNavigator
offers various options for results interpretation, e. g.
tables, graphs, bubble maps, contours, 1D and 2D histograms, well profiles, and various reports. The GUI provides an unlimited number of windows with different
types of data synchronized for detailed results monitoring.
Available for cluster as client-server application. All
simulation results for individual models or simulation
queues can be monitored in real time from a remote
terminal.
Dramatic boost to cluster usage efficiency!
Hydraulic
Fracture Modeling
tNavigator introduces a new approach to hydraulic
fracture simulation. The fracture is considered as part
of the well. To simulate fractures, a network of new “virtual” perforations is generated in the grid blocks intersected by the assumed fracture surfaces. The fracture
efficiency is modeled through the individual virtual
perforation efficiencies and proppant properties (SPE
138071).
This approach provides the most realistic calculation of fluid inflow to the well. This new technology
was successfully tested on large full-field models of giant Western Siberia reservoirs with a large number of
hydraulic fractures.
degree of flexibility in the definition and re-configuration of fracture networks. The orientation and shape of
fractures can be easily changed by modification of the
few keywords in the model stack file.
The new technology of fracture modeling can be
easily extended from single to multiple fractures per
well. The approach is found to be very efficient for simulation of shale reservoirs with multistage fractured
wells (SPE 147021).
Multiple hydraulic fractures changing pressure distribution
Hydraulic fracturing in tNavigator can be defined
quickly via the graphical interface. User defines fracture parameters in the dialogue (azimuth angle, length,
width, proppant type etc.). The fracture is immediately
visualized in the 3D map as a plane. Fracture geometry
can be any shape.
The main benefits of such an approach include a
more realistic description of fracture physics featuring more direct coupling of the well and the parts of
the reservoir connected through the fracture. This approach offers a simple way to define various complex
fracture properties as a function of time, and a high
Multiple realizations of fractures and reservoir properties can be applied for uncertainty quantification of
the reservoir performance.
Parallel Interactive Reservoir Simulations
Well trajectory
optimization
• Well path definition with a mouse.
• Run multiple simulations and compare results obtained with different sets of trajectories and input
parameters.
• Well trajectories can be visualized directly on the
maps.
Sector modeling tools.
Automatic model split/merge
This unique feature in tNavigator provides the user
with tools for automatic splitting of large models with
single and dual porosity/permeability into self-consistent sector models and merging them back together
again after updates. The input decks for sector models
are generated automatically, and include all features
and keywords from the initial model applied to the
small model grid.
• Efficient framework for project teams.
• Reduced overall simulation time.
• Simplified history matching and production optimization problems on sector models.
• Reduced project time frames.
• No painful data preparation work.
Users can define several alternatives for horizontal
well trajectories on 2D and 3D maps and user-defined
vertical profiles or cross-sections.
• tNavigator also allows direct definition of well parameters by including trajectories, events, history
in the DATA-file.
• No well section revision is required when you
change the grid. As the well definitions are given
in absolute coordinates it is common for all realizations of the model.
• Production optimization applications can change
the trajectories and well controls directly in the
tables without any additional scripting.
The model can be split using any trajectory: a specific region of the model, or via an arbitrary user map.
After running the large model one time before splitting, the flux boundary conditions are applied to all
fragments and used in the calculations. Model updates
recorded in the fragments are then merged back into
the large model.
The technology was applied to a number of giant
models in West Siberia (SPE 162090).
Advanced waterflood
optimization
• Add 2D, 3D streamline visualization and analytics,
while running full finite difference dynamic reservoir simulations.
Runtime graph of injection efficiency for all the injector wells
in the model. The amount of water injected — X axis, effective
oil production rate associated with a given injector — Y
• Generate tables of injector-producer cross couplings
and respective water, oil production allocation factors.
• Introduce tracers interactively to mark the water
component based on its origin. See how water front
moves from the injectors and examine the reservoir
water flow explicitly with maps and production plots.
Tracer maps and plots
Runtime table of injector-producer couplings calculated via
3D streamlines
• Interactively adjust the production plan: shut
wells, change the injection/production controls,
switch wells from production to injection, add new
wells, well patterns.
• Restart the run with a new production plan interactively, see the changes in the production rates and
their effective allocations, pressure distribution,
and watercuts.
• Check for underperforming injectors in the injec­
tion efficiency graphs.
Production profiles along the well completion intervals
Parallel Interactive Reservoir Simulations
Thermal modeling
Modeling thermal processes has posed many challenges as it requires very fine grid systems to capture
the heat transfer phenomena, the heterogeneity of porous media, and the complexity of fluid and heat flow
in multi-segmented wellbores.
The challenges in the simulation of thermal models
have been addressed in tNavigator. The advancement
in software and hardware technologies delivers the scalability of several orders of magnitude as compared to
single core simulation and the results are benchmarked
against industry simulators. The capability of handling
finer scaled models within a practical time frame allows
runningthe simulation with more precise prediction of
thermal processes and oil production. This opens the
door to tackling much larger and more detailed reservoir model resolution up to geological scale.
Thermal models such as steam-assisted gravity
drainage (SAGD), steam flooding and cyclic steam injection can be simulated with tNavigator.
tNavigator is a full reservoir thermal simulator with
capability of handling:
•
•
•
•
•
•
•
Dual porosity/ Dual permeability
Steam control for wells
Thermal aquifers
Streamline visualization
Electrical heaters
Multi-segmented wellbores
E300 and STARS input keywords
EOR modeling
tools
tNavigator proposes an elegant method for chemical injection simulations based on active tracers. The
particles of chemicals in the water phase can be fully
traced in the reservoir from the injection date, and can
induce changes in reservoir properties when the activation conditions are fulfilled. Based on user defined
tables the tracers can change reservoir properties, such
as permeability, porosity, relative permeability etc. The
model can be applied directly to the data available from
the laboratory core measurements.
tNavigator has been successfully applied to projects
with various EOR methods:
•
•
•
•
Alkalines
Surfactants
Polymers
BrightWater® nano-polymers
A similar mathematical approach is applied to waterflood problems, where the injected water salinity is
different from formation water salinity. A new mathematical model is proposed, taking into account basic
physical parameters of the undergoing process, such as
maximal salt concentration (saturated solution concentration), viscosity and water density dependence on
salt concentration, porosity, and permeability changes
with fresh water flooding (SPE 162091).
Full modeling workflow with tNavigator®
All modules are implemented on the same platform. Easy setup for
integrated workflows involving static and dynamic modeling
Uncertainty quantification
and assisted history matching
• Internal queuing system for distributing multiple
simulation runs locally or on remote clusters
• Various experimental design methods for sensitivity analysis
• Iterative algorithms for automatic history matching and production optimization
• Analysis toolbox: plots, histograms, tornado charts,
correlation analysis
• Direct variable definitions in the data files
• Workflow wizards
• Well trajectories, perforation intervals, fracture
geometry can be adjusted directly
• Variable LGR’s along well paths and fractures
• Corey functions for relative permeability modifications
• Reservoir property correlations with keywords
• Flexible user arithmetic for complex cases — no
external scripting required
• Local adjustments on sector models with flux
boundary conditions
Geological
modeling
• Software for building static models from scratch
• Using tNavigator platform for parallel computations and visualizations of large volume of data
• Very fast even on large models
• User friendly intuitive workflow manager
• Loading well trajectories, markers, horizons and
well log data in conventional formats
• Creating mesh using seismic trends
• Various methods for reservoir property interpolation from the well logs:
• 3D interpolation of reservoir maps
• Kriging
• Sequential Gaussian simulation method
• Flexible reservoir property calculator
• Export to dynamic model
• Basic correlations of fluid properties
• Corey functions for relative permeability parameters
• Standing’s PVT correlations
“The tNavigator technology represents a game-changer for us compared to other
reservoir simulation software in our organization. We not only can tackle far more
complex reservoir models with the software, but we are also able to fully exploit the
exceptional speed of tNavigator in combination with our assisted history matching
software to significantly reduce project cycle times. This in turn has made reservoir
simulation a much more valuable tool to our organization.”
- Larry Murray. Manager, Waterflood Modeling, Occidental Oil and Gas California Operations
“Having been a user of reservoir simulation for over 30 years, I was looking for the
next step change in technology that allowed us to do the things we wanted to do, at
the resolution we wanted, in an acceptable timeframe and at reasonable cost. RFD
achieved this, and I’m sure will achieve a significant part of the reservoir simulation
marketplace as others realise that this is a step change in the performance/price
value driver.”
- Steve Flew. Technical Director, Petrofac Malaysia
“I would like to express my appreciation for RFD’s outstanding support and thank
your development team for adding surfactant modeling capabilities to tNavigator.
Use of this feature in our simulation work has advanced our understanding of surfactant flooding potential in the Spraberry oil field in West Texas.“
- oil and gas exploration and production company
“With tNavigator we managed to reduce the simulation time of the Samotlor field
model from one day to 39 minutes. It completely changes the approach to development planning and waterflood optimization of this huge brown field.”
“RFD reduced the complexity of routine operations with the models to the general
PC user level”
- Vladislav Dzyuba. Director of Reservoir Modeling, TNK-BP
Website: www.rfdyn.com
Email: info-rfd@rfdyn.com
• MOSCOW • HOUSTON • KL • LONDON • CALGARY • BEIJING • HO CHI MINH CITY • MUMBAI • PERTH •
Download