Robust Optimal Control for Nonlinear Dynamic Systems Moritz Diehl, - - PowerPoint PPT Presentation

▶
robust optimal control for nonlinear dynamic systems
SMART_READER_LITE
LIVE PREVIEW

Robust Optimal Control for Nonlinear Dynamic Systems Moritz Diehl, - - PowerPoint PPT Presentation

Dec 08, 2022 27 likes •514 views

Robust Optimal Control for Nonlinear Dynamic Systems Moritz Diehl, Professor for Optimization in Engineering, Electrical Engineering Department (ESAT) K.U. Leuven, Belgium Joint work with Peter Kuehl, Boris Houska, Andreas Ilzhoefer

slide-1
SLIDE 1

Moritz Diehl

Robust Optimal Control for Nonlinear Dynamic Systems

Moritz Diehl, Professor for Optimization in Engineering, Electrical Engineering Department (ESAT) K.U. Leuven, Belgium Joint work with Peter Kuehl, Boris Houska, Andreas Ilzhoefer

INRIA-Rocquencourt, May 31, 2007

slide-2
SLIDE 2

Moritz Diehl

Overview

Dynamic Optimization Example: Control of Batch Reactors How to Solve Dynamic Optimization Problems? (recalled) Two Challenging Applications:

  • Robust Open-Loop Control of Batch Reactor
  • Periodic and Robust Optimization for „Flying Windmills“
slide-3
SLIDE 3

Moritz Diehl

Overview

Dynamic Optimization Example: Control of Batch Reactors How to Solve Dynamic Optimization Problems? Two Challenging Applications:

  • Robust Open-Loop Control of Batch Reactor
  • Periodic and Robust Optimization for „Flying Windmills“
slide-4
SLIDE 4

Control of Exothermic Batch Reactors

Cooperation between Heidelberg University and Warsaw University of Technology

Work of Peter Kühl (H.G. Bock, Heidelberg) with A. Milewska, E. Molga (Warsaw)

slide-5
SLIDE 5

Moritz Diehl

Batch Reactor in Warsaw [Peter Kuehl, Aleksandra Milewska]

Esterification of 2-Butanol (B) by propionic anhydride (A): exothermic reaction, fed-batch reactor with cooling jacket Aim: complete conversion of B, avoid explosion! Control: dosing rate of A

slide-6
SLIDE 6

Moritz Diehl

Safety Risk: Thermal Runaways

accumulation - temperature rise - thermal runaway

Try to avoid by requiring upper bounds on

  • reactor temperature TR , and hypothetical
  • adiabatic temperature S that would result if all A reacts with B
slide-7
SLIDE 7

Moritz Diehl

Differential (Algebraic) Equation Model

(1) (2)

slide-8
SLIDE 8

Moritz Diehl

Dynamic Optimization Problem for Batch Reactor

Constrained optimal control problem: Generic optimal control problem:

minimize remaining B subject to dosing rate and temperature constraints

slide-9
SLIDE 9

Moritz Diehl

Overview

Dynamic Optimization Example: Control of Batch Reactors How to Solve Dynamic Optimization Problems? Three Challenging Applications:

  • Robust Open-Loop Control of Batch Reactor
  • Periodic and Robust Optimization for „Flying Windmills“
slide-10
SLIDE 10

Moritz Diehl

Recall: Direct Multiple Shooting [Bock, Plitt 1984]

slide-11
SLIDE 11

Moritz Diehl

Solution of Peter’s Batch Reactor Problem

slide-12
SLIDE 12

Moritz Diehl

Experimental Results for Batch Reactor

Mettler-Toledo test reactor R1 batch time: 1 h end volume: ca. 2 l

slide-13
SLIDE 13

Moritz Diehl

Experimental Results for Batch Reactor (Red)

large model plant mismatch Safety critical! How can we make Peter and Aleksandra‘s work safer?

Blue: Simulation Red: Experiments

slide-14
SLIDE 14

Moritz Diehl

Experimental Results for Batch Reactor (Red)

large model plant mismatch Safety critical! How can we make Peter ‘s and Aleksandra‘s work safer?

slide-15
SLIDE 15

Moritz Diehl

Overview

Dynamic Optimization Example: Control of Batch Reactors How to Solve Dynamic Optimization Problems? Two Challenging Applications:

  • Robust Open-Loop Control of Batch Reactor
  • Periodic and Robust Optimization for „Flying Windmills“
slide-16
SLIDE 16

Moritz Diehl

Robust Worst Case Formulation

Make sure safety critical constraints are satisfied for all possible parameters p! Semi-infinite optimization problem, difficult to tackle...

slide-17
SLIDE 17

Moritz Diehl

Approximate Robust Formulation[Körkel, D., Bock, Kostina 04, 05]

Intelligent safety margins (influenced by controls) Fortunately, it is easy to show that up to first order:

~ ~

So we can approximate robust problem by:

slide-18
SLIDE 18

Moritz Diehl

Numerical Issues for Robust Approach

for optimization, need further derivatives of treat second order derivatives by internal numerical differentiation in ODE/DAE solver implemented in MUSCOD-II Robust -Framework [C. Kirches] use homotopy: start with nominal solution, increase slowly, employ warm starts

slide-19
SLIDE 19

Moritz Diehl

Estimated Parameter Uncertainties for Test Reactor

Standard deviation gamma Tjacket 0.3 K 3.0 mcatalyst 0.5 g (~10 %) 3.0 UA 10.0 W/(m2 K) (~10 %) 2.0 uoffset 5.0 10-5 kg/s (~10 % of upper bound) 3.0

slide-20
SLIDE 20

Moritz Diehl

Robust Open Loop Control Experiments

slide-21
SLIDE 21

Moritz Diehl

Robust Optimization Result and Experimental Test

Safety margin Perturbed Scenarios (Simulated) Blue: Simulation Red: Experiments

slide-22
SLIDE 22

Moritz Diehl

Comparison Nominal and Robust Optimization

Different solution structure. Model plant mismatch and runaway risk considerably

  • reduced. Complete conversion.
slide-23
SLIDE 23

Moritz Diehl

Comparison Nominal and Robust Optimization

Different solution structure. Model plant mismatch and runaway risk considerably

  • reduced. Complete conversion.
slide-24
SLIDE 24

Moritz Diehl

Overview

Dynamic Optimization Example: Control of Batch Reactors How to Solve Dynamic Optimization Problems? Two Challenging Applications:

  • Robust Open-Loop Control of Batch Reactor
  • Periodic and Robust Optimization for „Flying Windmills“
slide-25
SLIDE 25

Moritz Diehl

Conventional Wind Turbines

Due to high speed, wing tips are most efficient part of wing High torques at wings and mast limit size and height of wind turbines But best winds are in high altitudes! Could we construct a wind turbine with only wing tips and generator?

slide-26
SLIDE 26

Moritz Diehl

Conventional Wind Turbines

Due to high speed, wing tips are most efficient part of wing High torques at wings and mast limit size and height of wind turbines But best winds are in high altitudes! Could we construct a wind turbine with only wing tips and generator?

slide-27
SLIDE 27

Moritz Diehl

Crosswind Kite Power (Loyd 1980)

use kite with high lift-to-drag-ratio use strong line, but no mast and basement automatic control keeps kites looping But where could a generator be driven?

slide-28
SLIDE 28

Moritz Diehl

New Power Generating Cycle

New cycle consists of two phases: Power generation phase:

  • add slow downwind motion by

prolonging line (1/3 of wind speed)

  • generator at ground produces

power due to large pulling force Retraction phase:

  • change kite‘s angle of attack to

reduce pulling force

  • pull back line

Cycle produces same power as (hypothetical) turbine of same size!

slide-29
SLIDE 29

Moritz Diehl

New Power Generating Cycle

New cycle consists of two phases: Power generation phase:

  • add slow downwind motion by

prolonging line (1/3 of wind speed)

  • generator at ground produces

power due to large pulling force Retraction phase:

  • change kite‘s angle of attack to

reduce pulling force

  • pull back line

Cycle produces same average power as wind turbine of same wing size, but much larger units possible (independently patented by Ockels, Ippolito/Milanese, D.)

slide-30
SLIDE 30

Moritz Diehl

Can stack kites, can use on sea

slide-31
SLIDE 31

Moritz Diehl

Periodic Optimal Control (with Boris Houska)

forces at kite Control inputs: line length roll angle (as for toy kites) lift coefficient (pitch angle) ODE Model with 12 states and 3 controls Have to regard also cable elasticity

slide-32
SLIDE 32

Moritz Diehl

Some Kite Parameters

e.g. 10 m x 50 m, like Boeing wing, but much lighter material standard wind velocity for nominal power of wind turbines

slide-33
SLIDE 33

Moritz Diehl

Solution of Periodic Optimization Problem

Maximize mean power production: by varying line thickness, period duration, controls, subject to periodicity and other constraints:

slide-34
SLIDE 34

Moritz Diehl

Solution of Periodic Optimization Problem

Maximize mean power production: by varying line thickness, period duration, controls, subject to periodicity and other constraints:

slide-35
SLIDE 35

Moritz Diehl

Visualization of Periodic Solution

slide-36
SLIDE 36

Moritz Diehl

Prototypes built by Partners in Torino and Delft

New cycle consists of two phases: Power generation phase:

  • add slow downwind

motion by prolonging line (1/3 of wind speed)

  • generator at ground

produces power due to large pulling force Pull back phase:

  • change kite‘s angle of attack

to reduce pulling force

  • pull back line

Cycle allows same power production as wind turbine of same size!

slide-37
SLIDE 37

Moritz Diehl

Experimental Proof of Concept in Italy

slide-38
SLIDE 38

Moritz Diehl

What about ‚dancing‘ kites ?

slide-39
SLIDE 39

Moritz Diehl

  • 2 x 500 m2 airfoils
  • kevlar line 1500 m, diameter 8 cm
  • wind speed 10 m/s

Optimization with ‚dancing‘ kites: 14 MW possible

slide-40
SLIDE 40

Moritz Diehl

Question: could kite also fly without feedback? Stability just by smart choice of open-loop controls?

slide-41
SLIDE 41

Moritz Diehl

Linearization of Poincare Map determines stability

„Monodromy matrix“ = linearization of Poincare Map.

Stability Spectral radius smaller than one. Cons of Spectral radius:

  • Nonsmooth criterion difficult for optimization
  • uncertainty of parameters not taken into account
slide-42
SLIDE 42

Moritz Diehl

Periodic Lyapunov Equations and Stability

Lyapunov Lemma [Kalman 1960]: Nonlinear system is stable

  • periodic Lyapunov Equation

with has bounded solution.

slide-43
SLIDE 43

Moritz Diehl

Robust stability optimization problem Allows to robustly satisfy inequality constraints!

slide-44
SLIDE 44

Moritz Diehl

Orbit optimized for stability (using periodic Lyapunov eq.)

Open-loop stability only possible due to nonlinearity!

Long term simulation: Kite does not touch ground

slide-45
SLIDE 45

Moritz Diehl

Alternative: NMPC Control after turn of wind direction

slide-46
SLIDE 46

Moritz Diehl

Summary: Nonlinear Dynamic Optimization

Open-Loop Optimization (prone to model-plant-mismatch) Robust Open-Loop (no sensor feedback needed, simple) Model Predictive Control (feedback by fast online

  • ptimization)
slide-47
SLIDE 47

Moritz Diehl

Two events of interest this year

Workshop on NMPC Software and Applications (NMPC-SOFAP), Loughborough, United Kingdom, April 19-20, 2007.

(inv. speakers: Biegler, Findeisen, Kerrigan, Richalet, Schei)

13th Czech-French-German Conference on Optimization (CFG07), Heidelberg, Germany, September 17-21, 2007.

(inv. speakers: Fletcher, Scherer, Trelat, Waechter,...) Traditionally strong in optimal control.

slide-48
SLIDE 48

Moritz Diehl

4 PhD Positions in Numerical Optimization:

  • Sequential Convex Programming Algorithms for Nonlinear SDP
  • Large Scale & PDE Constrained Real-Time Optimization Algorithms
  • Fast Model Predictive Control Applications in Mechatronic Systems
  • Shape Optimization of Mechanical Parts under Inertia Loading

(deadline: June 21, 2007)