Integrated Optimization Software Issues · Integrated Optimization Software Issues Curtis H....
Transcript of Integrated Optimization Software Issues · Integrated Optimization Software Issues Curtis H....
Integrated Optimization Software Issues
Curtis H. WhitsonProfessor NTNU
Dept. of Petroleum Engineering & Applied Geoscience
Founder PERA & Petrostreamz
IO09 – Trondheim Sept. 29, 2009
2
Integrated Optimization
Holistic …
the value of a composite system is far greater than the sum of the elements making up the system.
the value of a composite system is far greater than the sum of the elements making up the system.
Schlumberger
To sell oil & gas we have to firstpipe together the operation…
3
Integrated Optimization
has a value greater than the sum of the elements making up the integrated system?
4
Integrated Optimization
has a value greater than the sum of the elements making up the integrated system?
allows ranking of control variables…
5
Integrated Optimization
has a value greater than the sum of the elements making up the integrated system?
allows ranking of control variables…
based on bottomline
economics.
6
Software High-Level Issues
• Models• Integration• Optimization• IT• GUI• Costs• Sustainability
Schlumberger
14
No Integration – Silo Modeling Approach
Holistic …
the value of a composite system is far greater than the sum of the elements making up the system.
the value of a composite system is far greater than the sum of the elements making up the system.
Schlumberger
Reservoir
Pipeflow
Process Economics
15
Software Models
• Use current models & workflow.– Allow continued “silo” modeling.– Allow any app – ecumenical.– Platform independent.
• Allow multiple surrogates & proxies.
16
Today’s Modeling
the value of a composite system is far greater than the sum of the elements making up the system.
Schlumberger
Reservoir
Pipeflow
Process Economics
PipeFlowGAPOLGA
ECLSensorMatbal
HysysUnisysNews
MerakAiresPetec
Linux
Cluster
UnixLaptop
17
Model
Make things
as simple as possible, but
not simpler...Albert Einstein
A mathematical representation of a physical process.
18
Model Types• Theoretical.
– Based on rigorous physical laws.• Transport.• Conservation.• Equilibrium
• Empirical Proxies.– Based on simplified physical laws.– Correlation of data.– Best-fit to more-rigorous models.
19
Modeling Issues
• Cost, availability & license management.
• Computational requirement.– Solution methods.– CPU & memory efficiency.
• Model Parameters.– Estimation.– Tuning to observations.
20
• Use current models & workflow.• Proprietary vs Open.• Platform dependent.• System compliant.• Multiple surrogates & proxies.
Software Model Surrogates & Proxies
Compositional – 3 hr
Black-Oil – 30 min
Streamline – 3 min
Material BalanceIPR – 3 sec
21
A vision of Integrated Modeling
PipeFlowGAPOLGA
HysysUnisysNews
ECLSensorMatbal
MerakAiresPetec
Schlumberger
22
Software Integration
• Automatic model orchestration.• Flexible data integration.• Modularity & reusability.• Multi-level integration.• Global & local integration.• Nesting, branching, looping.
23
Integration Issues
• Degree of coupling.• Order dependence.• Parallelization.• Re-execution logic.• Data linking.• Stream translations.
24
A vision of Integrated Optimization
PipeFlowGAPOLGA
HysysUnisysNews
ECLSensorMatbal
MerakAiresPetec
Schlumberger
25
Software Optimization
• Optimizer-Model interaction.• Defining optimization parameters.
– Target, variables, constraints.• Solver library.• Nested optimizations.• Defining cost functions.
26
Optimization Issues• Target Objective (O)
– Minimization, maximization,feasibility– Weight factors.– Scaling.
• Variables (V)– Real and integer.
• Constraints (C)• Derivatives
– dO/dV & dC/dV.
27
Optimization Issues
• Solvers– Derivative based.– Non-derivative based.– Mixed integer non-linear (MINL).– Experimental design.– Manual cause-and-effect case matrices.– Nested optimizations, different solvers.– Proprietary vs open.
28
Software IT
• Use all available resources.• Flexible data access and transfer.• Handle OS handshaking.• Localization, storage & backup.
29
Software GUI
• Platform independent.• Visualization of elements.• Modular visualization.• Multi-level visualization (V & H).• Intuitive and user friendly.• Embedded model GUIs.
30
Software Costs
• Licensing & in-house development.• IO Project manager.• IT engineers – data interfacing.• Training.• Cross-disciplinary collaboration.