Master
thesis topics 2010
The text
below is a very short description of the topics. Contact Lars Imsland
for further information. I
am also open for suggestions.
1.
Modellbasert
regulering og optimalisering med store fysikalske modeller.
Konkret case vil være å se på total produksjonsoptimalisering av offshore prosessanlegg. Hele produksjonskjeden fra brønn til eksport skal inkluderes: Produksjonsanlegg (brønner, rørledninger etc) og topside prosess (separatorer, kompressorer, vannbehandling). En nokså komplett modell i Modelica finnes fra før.
Deloppgaver er:
· Utvikling av ulineær MPC for aktuell prosess (videreutvikle eksisterende modell)
· Simulering av anbefalt løsning
· Metodestudie
Medveileder: Svein Olav Hauger, Pål
Kittilsen, Cybernetica AS
2. Implementation
and comparison of methods for sensitivity integration for gradient calculation
in non-linear MPC
Practical implementation of analytical
gradient-computation for 'single-shooting' non-linear MPC is usually based on
sensitivity integration. The candidate should implement (at least) two
different methods for sensitivity integration (BDF and the algorithm of
Schlegel et al (2004)), and assess their usefulness and appropriateness for
different types of NMPC problems.
3. Alternative
formulations for fast non-linear MPC
The candidate should explore implementation of
nonlinear MPC using l1-penalty
functions for constraints. This has the advantage of very fast gradient
computation, but on the other side the objective function becomes non-smooth.
Co-supervisor: Ruben Ringset, Cybernetica AS
4. Parallellization in
non-linear MPC
Some types of implementation of nonlinear MPC can
benefit from parallellization. The candidate should
explore strategies for parallellization, and make
prototype implementations for assessment and comparison.
5. Implementation
of framework for non-linear MPC based on open-source components
The candidate should investigate open-source
components that fit into a framework for non-linear MPC. Relevant components
might be for modeling, code-generation, and optimization.
6. Frequency
tracking/wave spectrum estimation based on Lyapunov-theory
and non-linear adaptation.
The candidate should implement and analyze a
non-linear adaptation law for frequency tracking and/or wave spectrum
estimation based on a specific parameterization of harmonics.