Robust control of resistive wall modes using pseudospectra

Similar documents
Three Dimensional Effects in Tokamaks How Tokamaks Can Benefit From Stellarator Research

A NONSMOOTH, NONCONVEX OPTIMIZATION APPROACH TO ROBUST STABILIZATION BY STATIC OUTPUT FEEDBACK AND LOW-ORDER CONTROLLERS

Resistive Wall Mode Control in DIII-D

Non-linear MHD Simulations of Edge Localized Modes in ASDEX Upgrade. Matthias Hölzl, Isabel Krebs, Karl Lackner, Sibylle Günter

The Effects of Noise and Time Delay on RWM Feedback System Performance

Effects of Noise in Time Dependent RWM Feedback Simulations

Optimization based robust control

NORMS ON SPACE OF MATRICES

RWM Control in FIRE and ITER

Pseudospectra and Nonnormal Dynamical Systems

On existence of resistive magnetohydrodynamic equilibria

RWM FEEDBACK STABILIZATION IN DIII D: EXPERIMENT-THEORY COMPARISONS AND IMPLICATIONS FOR ITER

Modelling of wall currents excited by plasma wall-touching kink and vertical modes during a tokamak disruption, with application to ITER

Plasma Response Control Using Advanced Feedback Techniques

Model based optimization and estimation of the field map during the breakdown phase in the ITER tokamak

RESISTIVE WALL MODE STABILIZATION RESEARCH ON DIII D STATUS AND RECENT RESULTS

MHD limits and plasma response in high beta hybrid operations in ASDEX Upgrade

Analysis and modelling of MHD instabilities in DIII-D plasmas for the ITER mission

FIXED POINT ITERATIONS

Securing High β N JT-60SA Operational Space by MHD Stability and Active Control Modelling

Singular Value Decomposition of the frequency response operator

- Effect of Stochastic Field and Resonant Magnetic Perturbation on Global MHD Fluctuation -

Resistive Wall Mode Observation and Control in ITER-Relevant Plasmas

Introduction to Fusion Physics

RWM Control Code Maturity

A method for calculating active feedback system to provide vertical position control of plasma in a tokamak

High-m Multiple Tearing Modes in Tokamaks: MHD Turbulence Generation, Interaction with the Internal Kink and Sheared Flows

Linear Algebra Massoud Malek

Requirements for Active Resistive Wall Mode (RWM) Feedback Control

Nonlinear equations. Norms for R n. Convergence orders for iterative methods

Numerical Methods for Differential Equations Mathematical and Computational Tools

Non-linear modeling of the Edge Localized Mode control by Resonant Magnetic Perturbations in ASDEX Upgrade

GA A27857 IMPACT OF PLASMA RESPONSE ON RMP ELM SUPPRESSION IN DIII-D

1 Sylvester equations

Analytical Study of RWM Feedback Stabilisation with Application to ITER

An Iteration-Domain Filter for Controlling Transient Growth in Iterative Learning Control

Model based estimation of the Eddy currents for the ITER tokamak *

Flow Instability Matlab Tutorial Bangalore, January 2010 Matthew Juniper, Dan Henningson, Peter Schmid 1

Evaluation of Chebyshev pseudospectral methods for third order differential equations

Rotation and Neoclassical Ripple Transport in ITER

On the transient behaviour of stable linear systems

Evaluation of CT injection to RFP for performance improvement and reconnection studies

Upon successful completion of MATH 220, the student will be able to:

Global Mode Control and Stabilization for Disruption Avoidance in High-β NSTX Plasmas *

Course Summary Math 211

Numerical Computation of Structured Complex Stability Radii of Large-Scale Matrices and Pencils

ASSESSMENT AND MODELING OF INDUCTIVE AND NON-INDUCTIVE SCENARIOS FOR ITER

Gramians based model reduction for hybrid switched systems

Characterization of the Perpendicular Rotation Velocity of the Turbulence by Reflectometry in the Stellarator TJ-II

VISUALIZING PSEUDOSPECTRA FOR POLYNOMIAL EIGENVALUE PROBLEMS. Adéla Klimentová *, Michael Šebek ** Czech Technical University in Prague

Magnetic Flux Surface Measurements at Wendelstein 7-X

Fast Linear Iterations for Distributed Averaging 1

MATH 581D FINAL EXAM Autumn December 12, 2016

Princeton Plasma Physics Laboratory. Multi-mode analysis of RWM feedback with the NMA Code

Strong stability of neutral equations with dependent delays

MHD. Jeff Freidberg MIT

The performance of improved H-modes at ASDEX Upgrade and projection to ITER

Performance limits. Ben Dudson. 24 th February Department of Physics, University of York, Heslington, York YO10 5DD, UK

Critical Physics Issues for DEMO

The Kernel Trick, Gram Matrices, and Feature Extraction. CS6787 Lecture 4 Fall 2017

ROBUST STABLE NONLINEAR CONTROL AND DESIGN OF A CSTR IN A LARGE OPERATING RANGE. Johannes Gerhard, Martin Mönnigmann, Wolfgang Marquardt

MHD Induced Fast-Ion Losses in ASDEX Upgrade

Course Notes: Week 1

Derivation of dynamo current drive in a closed current volume and stable current sustainment in the HIT SI experiment

DIII D. by F. Turco 1. New York, January 23 rd, 2015

DIAGNOSTICS FOR ADVANCED TOKAMAK RESEARCH

Predictive Study on High Performance Modes of Operation in HL-2A 1

Mitigation of ELMs and Disruptions by Pellet Injection

Feedback stabilization of the resistive shell mode in a tokamak fusion reactor

Modeling of resistive wall mode and its control in experiments and ITER a

A Lagrangian approach to the kinematic dynamo

Perturbation Theory for Self-Adjoint Operators in Krein spaces

Section 3.9. Matrix Norm

Phase ramping and modulation of reflectometer signals

Supported by. Role of plasma edge in global stability and control*

Extended Lumped Parameter Model of Resistive Wall Mode and The Effective Self-Inductance

NIMROD FROM THE CUSTOMER S PERSPECTIVE MING CHU. General Atomics. Nimrod Project Review Meeting July 21 22, 1997

Analysis of Transient Behavior in Linear Differential Algebraic Equations

STABILIZATION OF m=2/n=1 TEARING MODES BY ELECTRON CYCLOTRON CURRENT DRIVE IN THE DIII D TOKAMAK

Feedback Control Systems

Performance, Heating, and Current Drive Scenarios of ASDEX Upgrade Advanced Tokamak Discharges

Beam Driven Alfvén Eigenmodes and Fast Ion Transport in the DIII-D and ASDEX Upgrade (AUG) Tokamaks

AN ELEMENTARY PROOF OF THE SPECTRAL RADIUS FORMULA FOR MATRICES

April 13, We now extend the structure of the horseshoe to more general kinds of invariant. x (v) λ n v.

Effect of Resonant and Non-resonant Magnetic Braking on Error Field Tolerance in High Beta Plasmas

Iterative Solution of a Matrix Riccati Equation Arising in Stochastic Control

Analysis of different Lyapunov function constructions for interconnected hybrid systems

B5.6 Nonlinear Systems

Global particle-in-cell simulations of Alfvénic modes

Stabilization of Distributed Parameter Systems by State Feedback with Positivity Constraints

Finding normalized and modularity cuts by spectral clustering. Ljubjana 2010, October

An efficient algorithm to compute the real perturbation values of a matrix

Computational Linear Algebra

Multi-Robotic Systems

GA A26247 EFFECT OF RESONANT AND NONRESONANT MAGNETIC BRAKING ON ERROR FIELD TOLERANCE IN HIGH BETA PLASMAS

Modeling of active control of external magnetohydrodynamic instabilities*

Properties of Matrices and Operations on Matrices

BOUNDS OF MODULUS OF EIGENVALUES BASED ON STEIN EQUATION

Plasma models for the design of the ITER PCS

arxiv: v1 [physics.plasm-ph] 24 Nov 2017

Transcription:

Robust control of resistive wall modes using pseudospectra M. Sempf, P. Merkel, E. Strumberger, C. Tichmann, and S. Günter Max-Planck-Institut für Plasmaphysik, EURATOM Association, Garching, Germany GOTiT Seminar, January 2009 Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 1 / 24

Outline 1 Code package for resistive wall mode feedback stabilization 2 Controller optimization: eigenvalues vs. pseudospectra 3 ITER-like test case 4 Conclusions Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 2 / 24

STARWALL code An external kink mode can be stabilized by an ideally conducting wall close to the plasma; if the wall is not superconducting, the mode grows on the resistive time scale of the wall resistive wall mode (RWM) STARWALL [1, 2]: 3D ideal MHD stability code specialized to RWMs (E kin neglected) feedback coil system included to stabilize RWMs coupling between different toroidal mode numbers n (3D effect) Inputs to STARWALL: plasma equilibrium 3D wall and coil geometries sensor positions and orientations feedback controller logics Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 3 / 24

General controller model Coils and sensors are grouped into toroidal arrays, respectively. Voltage vector applied to k-th coil array: s l G kl L R k i k L u k = G kl s l R k i k l=1 magnetic field perturbation vector measured by l-th sensor array proportional gain matrix linking coil array k to sensor array l number of sensor arrays artificial additional coil resistance (to make coils faster ) vector of coil currents in array k Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 4 / 24

Structure of the gain matrix G kl G kl is strucured in such a way that the coil array produces a field with toroidal mode number n in response to a perturbation with the same n: G kl ij = ( ) α kl n cos(nϕkl ij ) + βn kl sin(nϕkl ij ), n } {{ } in-phase response }{{} (90/n) phase-shifted response ϕ kl ij = toroidal angle between coil i of array k and sensor j of array l. The sum runs over all n s to be simultaneously controlled. Remaining free parameters defining the feedback logics: α kl n βn kl R k cosine gains sine gains additional coil resistances Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 5 / 24

Structure of the STARWALL equation Dynamics of the plasma-wall-coils system: Lẋ = Rx, R = R 0 + ( ) α kl n Rkl n,α + βkl n Rkl n,β n,k,l k R k Rk x L R 0 R kl n,α, R kl n,β, R k α kl n, β kl n, R k state vector (coil currents, wall current potentials) inductance matrix + plasma contribution resistance matrix matrices describing the effect of feedback free parameters x(t) e γt parametrized eigenvalue problem (R, L N N ): L 1 Rx i = γ i x i ; stability Re γ i < 0 i = 1,..., N Stabilizing parameter set OPTIM code [1, 3] Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 6 / 24

Automatic controller design procedure STARWALL compute matrices L, R 0, R kl n,α, Rkl n,β, R k TRANSFORM (model reduction) construct orthogonal pattern matrix P (details on next slide) transform each matrix M computed by STARWALL as M = P 1 MP in each M, retain only upper left N red N red block (N red N) OPTIM (feedback optimization in reduced system) find {α kl n, βkl n, R k } so that L 1 R is optimally stable (details later) Cross-check against full-sized system using the optimal set {α kl n, βkl n, R k }, compare properties of L 1 R with those of L 1 R (eigenvalues, pseudospectra,... ; details later) Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 7 / 24

Model reduction: isometric truncation The robust stability concept introduced later requires orthogonality of the projection underlying the model reduction = newly developed isometric truncation procedure: projection onto leading columns of the pattern matrix P Properties of the pattern matrix P = (p 1 p 2... p N ): each column p i represents a system state (current pattern) ohmic loss orthogonality: p T i R 0p j = δ ij, i = 1,..., N, j = 1,..., N first columns are unstable eigenvectors of L 1 R 0 (system without feedback), i., e., RWMs the remaining columns represent physical processes in the stable subspace of L 1 R 0, ordered according to decreasing controllability and observability Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 8 / 24

What does optimal stability mean? ẋ = Ax = x(t) = e ta x(0) = N ξ i x i e γ it i=1 here: A = L 1 R (asymptotic) stability: lim t x(t) = 0 Re γ i < 0 i = 1,..., N Basic objectives for stability optimization good asymptotic stability: Re γ i < q i = 1,..., N, q > 0 large robust stability: A stable = A + E stable for any moderate E With feedback, A is non-normal, i.e., the x i s are far from orthogonal. Exclusive features which non-normal matrices can have: extreme eigenvalue sensitivity (affects robustness) transient growth ( e ta 1 for some t > 0, although A is stable) another optimization objective: keep sup t>0 e ta small! Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 9 / 24

Measures of asymptotic stability The traditional measure of asymptotic stability: σ(a) = max i=1,...,n Re γ i (spectral abscissa) Minimization of σ(a) pushes the leading eigenvalue(s) as far as possible into the left complex halfplane. Another stability measure, seeing not only the leading eigenvalues: η(a) = N exp(re γ i ) i=1 ( exponential spectral function ) Minimization of η(a) pushes all the eigenvalues as far as possible to the left. However, neither σ(a) nor η(a) guarantee robustness of stability, because sensitivity of eigenvalues is not taken into account. Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 10 / 24

Eigenvalue sensitivity and robust stability An idea of sensitivity of the entire eigenvalue spectrum: γ ɛ (A) is the set of z such that z is an eigenvalue of A + E for some E N N with E < ɛ (ɛ-pseudospectrum) Measure for robustness of stability: ρ(a) = sup{ɛ : A + E is stable for all E N N with E < ɛ} (complex stability radius) Measure of stability for perturbations of maximum allowable ɛ : σ ɛ (A) = sup{re z : z γ ɛ (A)} (ɛ-pseudospectral abscissa) In particular, σ ρ(a) (A) = 0. Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 11 / 24

Relationship between σ ɛ (A) and e ta σ ɛ (A) and e ta are related by various theorems. One of them [4] : sup e ta σ ɛ (A)/ɛ ɛ > 0. t 0 That means, if the eigenvalues of A are so sensitive that σ ɛ (A)/ɛ > 1 for some ɛ, there must be transient growth. Tradeoff: optimize σ ɛ (A) for small ɛ = good asymptotic stability optimize σ ɛ (A) for large ɛ = temperate transient behavior Is optimization of ρ(a) a good compromise? Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 12 / 24

OPTIM s functionality Parallel eigenvalue optimization code for parametrized matrices A OPTIM s objective functions F 1 = σ(a) spectral abscissa F 2 = η(a) exponential spectral function F 3 = ρ(a) neg. complex stability radius ( two-step algorithm [5]) F 4 = σ ɛ (A) ɛ-pseudospectral abscissa ( criss-cross algorithm [6]) Minimization algorithm: gradient bundle method [7], suitable for non-smooth, non-lipschitz functions Additional features for given A: computation of ɛ-pseudospectra boundaries (contour plots in ) computation of e ta plots Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 13 / 24

ITER-like test case [3] plasma: Scenario 4 equilibrium, β = 2.29%, n = 1 perturbations only wall: interior wall only, simplified geometry coils: single array, 7 port plug coils (2 coils missing due to collision with NBI) sensors: single array, 18 sensors, z orientation 6 4 Z [m] 2 0 2 plasma sensor coil 4 wall 6 2 4 6 8 10 R [m] Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 14 / 24

Two unstable RWMs growth rates not exactly equal due to broken axisymmetry current patterns (φ isolines, j = n φ) almost equal, but toroidally phase-shifted by 90 RWM 1 (γ 1 = 21.9 s 1 ) RWM 2 (γ 2 = 21.7 s 1 ) Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 15 / 24

Model reduction: stable subspace patterns 1-4 Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 16 / 24

Model reduction: stable subspace patterns 5-8 Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 17 / 24

Model reduction: stable subspace patterns 9-12 Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 18 / 24

Stability optimization Only n = 1 plasma perturbations accounted for, single coil and sensor array = 3 free parameters: α, β, R Full model dimension: N = 5190; reduced model dimension: N red = 58 Optimization of F 1 = σ(a), F 2 = η(a), F 3 = ρ(a), F 4 = σ ɛ (A), where ɛ = 2ρ opt with ρ opt being the optimal value of ρ(a) obtained after optimizing F 3 Very good agreement between reduced and full model [3] Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 19 / 24

1 7 10 5 7 Pseudospectra after stability optimization boundaries of γ ɛ (A), ɛ indicated by contour labels, dots = eigenvalues angular frequency [1/s] angular frequency [1/s] 100 80 60 40 20 0 1.5 2.5 0.7 150 100 50 0 50 growth rate [1/s] 100 80 60 40 20 0 0.5 1 0.7 0.7 0.5 3.5 0.5 F 1, full model 150 100 50 0 50 growth rate [1/s] 1 1.5 0.5 0.7 2.5 5 3.5 5 7 15 F 3, full model angular frequency [1/s] angular frequency [1/s] 100 80 60 40 20 0 10.7 0.5 150 100 50 0 50 growth rate [1/s] 100 80 60 40 20 0.5 1 0.7 1 0.5 0.7 1.5 0.7 1 0.5 F 2, full model 0 150 100 50 0 50 growth rate [1/s] 1 3.5 2.5 1.5 2.5 7 3.5 5 F 4, full model 10 Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 20 / 24

Transient amplification ( e ta curves) e ta 160 120 80 F 1 F 2 F 3 F 4 40 0 0 0.25 0.5 0.75 1 t [10 1 s] Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 21 / 24

Conclusions Optimization of F 1 = σ(a) or F 2 = η(a) does not produce very robust stability, compared to optimization of F 3 = ρ(a) or F 3 = σ ɛ (A) Optimization of F 2 gives a catastrophic transient peak Even for the F 3 - and F 4 -optimal solutions, the transient behavior is not entirely satisfactory, but there are strategies to improve this further [3] Robustness and transient peaks might be an issue for ITER Robustness and transient behavior should generally be taken into account when designing and rating plasma scenarios to be RWM stabilized Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 22 / 24

Plans for the near future Improvement of the approximation used when taking a time delay between sensors and actuators into account Include the voltage loss in the busbar (this is simple) Realistic ITER modeling including the double wall with port extensions and blanket support, the blanket modules, and in-vessel coils AUG modeling Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 23 / 24

References P. Merkel and M. Sempf. Feedback stabilization of resistive wall modes in the presence of multiply-connected wall structures. 21st IAEA Fusion Energy Conference 2006, Chengdu, China, paper TH/P3-8, 2006. E. Strumberger, P. Merkel, M. Sempf, and S. Günter. On fully three-dimensional resistive wall mode and feedback stabilization studies. Phys. Plasmas, 15:056110, 2008. DOI:10.1063/1.2884579. M. Sempf, P. Merkel, E. Strumberger, C. Tichmann, and S. Günter. Robust control of resistive wall modes using pseudospectra. New J. Phys., 2009, submitted. L. N. Trefethen and M. Embree. Spectra and Pseudospectra - The Behavior of Nonnormal Matrices and Operators. Princeton University Press, Princeton and Oxford, 2005, 606 pp. N. A. Bruinsma and M. Steinbuch. A fast algorithm to compute the H -norm of a transfer function matrix. Syst. Contr. Lett., 14:287 293, 1990. J. V. Burke, A. S. Lewis, and M. L. Overton. Robust stability and a criss-cross algorithm for pseudospectra. IMA J. Numer. Anal., 23:359 375, 2003. J. V. Burke, A. S. Lewis, and M. L. Overton. Two numerical methods for optimizing matrix stability. Linear Algebra Appl., 351-352:117 145, 2002. Mario Sempf (IPP Garching) Robust RWM control using pseudospectra GOTiT Seminar, January 2009 24 / 24