1 / 14

Egemen “ Ege ” Kolemen , S. Gerhardt, D. Gates

NSTX. Supported by . Development of Fiducial Shots with LLD: Strike Point Control Improvement and Incorporation in Regular Operation. College W&M Columbia U Comp-X General Atomics INEL Johns Hopkins U LANL LLNL Lodestar MIT Nova Photonics New York U Old Dominion U ORNL PPPL PSI

garnet
Download Presentation

Egemen “ Ege ” Kolemen , S. Gerhardt, D. Gates

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. NSTX Supported by Development of Fiducial Shots with LLD: Strike Point Control Improvement and Incorporation in Regular Operation College W&M Columbia U Comp-X General Atomics INEL Johns Hopkins U LANL LLNL Lodestar MIT Nova Photonics New York U Old Dominion U ORNL PPPL PSI Princeton U Purdue U SNL Think Tank, Inc. UC Davis UC Irvine UCLA UCSD U Colorado U Maryland U Rochester U Washington U Wisconsin CulhamSciCtr U St. Andrews York U Chubu U Fukui U Hiroshima U Hyogo U Kyoto U Kyushu U Kyushu Tokai U NIFS Niigata U U Tokyo JAEA Hebrew U Ioffe Inst RRC Kurchatov Inst TRINITI KBSI KAIST POSTECH ASIPP ENEA, Frascati CEA, Cadarache IPP, Jülich IPP, Garching ASCR, Czech Rep U Quebec Egemen “Ege” Kolemen, S. Gerhardt, D. Gates 2009 NSTX XP Review, ASC Room B-318 January 5th, 2009

  2. This year NSTX will need a new fiducial shot development due to the new restrictions and requirements due to LLD installation. • We can’t let the strike point hit the LLD (at least not routinely) • Medium δ shots may be part of regular XPs. • Strike point controllers are operational but they are still not part of the regular run. • We need to improve and prove that the controller works robustly Need for Developing a New Fiducial with LLD

  3. Long term aim: • Use all the PF coils to control the plasma shape together. • Very hard to implement at once. • Incrementally increase the control capability to reach aim Full Multiple-Input-Multiple-Output Control -Inner Gap -Lower Inner Strike Point -Vertical Position -Squareness Upper X-point Height …

  4. System Id: Identify the effect of these coils on the boundary shape. • Last year: Reaction Curve Method • Results from last year: • Problem: • Many shots needed • Not precise Current System ID P

  5. Inner Strike Point Control on the Vertical Divertor via a PI controller with PF2L • PI tuned to [5000, 5000] for inner strike point. • RMS ~ 1cm (Has offset need higher I) Current Capability: Inner Strike Point Control PF1AL PF2L PF2L controls outer SP in red segments. PF1AL controls inner SP in the blue segment.

  6. Outer Strike Point Control on the Inner Divertor via a PI controller with PF2L • PI controller gains are tuned using Ziegler and Nichols to [400, 800]. • RMS ~ 1.5 cm (Kappa ~2.2-2.3) Current Capability: Outer Strike Point on the Inner Divertor PF1AL PF2L PF2L controls outer SP in red segments.

  7. Outer Strike Point Control on the Outer Divertor via a PI controller with PF2L • PI controller gains are tuned using Ziegler and Nichols to [400, 800]. • RMS ~ 2 cm (Has offset need higher I) Current Capability: Outer Strike Point on the Outer Divertor PF1AL PF2L PF2L controls outer SP in red segments. PF1AL controls inner SP in the blue segment.

  8. Is bull-nose tile addition a problem for OSP? • Not a big problem. PCS uses a linear segment to find the SP, we can’t make it a curved surface. Maybe the line segment can be modified but not really that important. SP control tries to match flux on two points, if the reference is moved down. • Present status of OSP control on the inner divertor. Do we have a shot? Do we need more work? • We have a shot, need to add D gain to stabilize. • Transient control compared to control during the Ip flat top. • This is the main problem. Strike point moves all around. Plasma touches the bottom of the vessel. With sprayed lithium on the inner divertor, this may be a problem. • Implement X-point height control. • Adding D to SP controller. • Upper strike point controllers Problems That Needs to be Addressed before LLD Commissioning

  9. This year Offline System ID (via Subspace Numerical Method): • Find minimal-state realization by converting to a form where error matrix (wk and vk) is minized • We parameterize the matrices by A(θ) and B(θ), then solve NLP: • Best known method: L. Ljung: ARMAX input-output models used in Maximum Likelihood. • IMPORTANT: System Delays are not included in this formalism. • Include that as additional parameters in the optimization. This Year: Offline System ID (Subspace Numerical Method)

  10. Requirements: • Need an equilibrium • Need variation of the control input (PF1/2…). • Thus a working controller (can be a very poor one) is needed. • Then clean up the response • Find the portion of the response that depends mostly on the control • Take out trend lines, biases… • Use Subspace Algorithm with delays to find system. • Pros: • Don’t need XP run time! • Great for improving a already existing controller. • Cons: • Precision may not be perfect. • Need a running controller Offline System ID (via Subspace Numerical Method)

  11. This year Offline System ID (via Subspace Numerical Method): • Example: System Id for PF2L to OSP compared to XP data. • Enables better initial guess within a factor of 2 • Need a working controller for best results. • Guesses Obtained: • I for PF3 • D for SP (PF1/PF2) • Off-diagonal terms for PF1/PF2 to ISP/OSP controller. • Initial guess for X-point controller harder (no working controller) • When we reach this closed-loop plant response pattern the oscillation period (Pu) and the amplitude (A) of the plant response can be measured and used for PID controller tuning. where • Only a single experiment is needed. • Closed loop: More stable • K= 2.4e-6, T = 9.9 and L = 6.0 • K =3.5e-6, T=10, and L =7 • tauD_method2 = • 0.010 • Relay Feedback is almost implemented on PCS. This Year: Offline System ID Note that shot used for system ID is not in this list

  12. We propose a 1 day XP: • Load shot 134986 and see if the shot is still the same and SP controllers are working (2 shots) • Improve control: We will be adding derivative gain and tuning the controllers for better performance (XMP): • Add D to PF2 for strike point control – Reduce RMS < 1 cm (3 shots) • Relay Feedback System ID • Multi-Input Multi-Output: PF1A/PF2 to ISP/OSP (3 shots) [May not work] • Add I to PF3 – Need to get rid of Bias Error (~1 cm) (3 shots) • Test Integral Fix (1 shot) • Add upper strike point controllers: Currently, the strike point controllers work only for the lower side, we will implement the upper strike point controller in PCS. (4 shots) • Scan the strike-point from 35-40-45-50-55-63 cm. (5 shots) • Improve transient phase: Study and tune the start phase to avoid the plasma touching the lower plasma boundary. (6 shots) Experimental Plan for the New Fiducial

  13. Study and tune the start phase to avoid the plasma touching the lower plasma boundary. • Implement the X-point height controller • After the Integrator start fix, try to start the controller later (i.e during flat top). Set the PF1/PF2 to constant values for the ramp up. • Tweak the squareness and Drsep. • Change upper strike point locations • Other ideas? Other Ways to Improve Transient Phase?

  14. Backup Slides

More Related