XP 1058: Impact of Outer Squareness on High-kappa Discharge XP 1058: Impact of Outer Squareness on High-kappa Discharge PerformancePerformance

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

2010 NSTX XP ReviewRoom B-318

July 30th, 2010

College W&MColumbia UComp-XGeneral AtomicsINELJohns Hopkins ULANLLLNLLodestarMITNova PhotonicsNew York UOld Dominion UORNLPPPLPSIPrinceton UPurdue USNLThink Tank, Inc.UC DavisUC IrvineUCLAUCSDU ColoradoU MarylandU RochesterU WashingtonU Wisconsin

Culham Sci CtrU St. Andrews

York UChubu UFukui U

Hiroshima UHyogo UKyoto U

Kyushu UKyushu Tokai U

NIFSNiigata UU Tokyo

JAEAHebrew UIoffe Inst

RRC Kurchatov InstTRINITI

KBSIKAIST

POSTECHASIPP

ENEA, FrascatiCEA, Cadarache

IPP, JülichIPP, Garching

ASCR, Czech RepU Quebec

NSTXNSTX Supported by

• Keep all the coil currents same other than PF5

• Change PF4 only

• PF5 compensates for PF4 by increasing 3/2*PF4 to keep the plasma in the vessel.

• ΔPF5 = -1.5*ΔPF4

PF4 Scan from -6 kA to +7 kA

Effect of PF4 without X-Point Control

PF4

• PF1 and PF2 stays roughly constant.

• PF5 shows the same -1.5*ΔPF4 behavior.

• Minimal compensation in PF3: ΔPF3 = 0.15*ΔPF4

• We can change squareness while keeping the other parameters cons

• Inner gap moves minimally (~1 mm) for 1kA change in PF4.

PF4 Scan from -10 kA to +10 kA

Effect of PF4 with X-point/Outer Gap Controller On

• Add a new segment to control squareness with PF4.

• Keep everything else the same for first XP.

• Move the segment for PF3 inwards to give more leeway to PF4 if needed.

• Once the control works, do a squareness scan.

Choose Segment Along the Highest Change Direction.

How to Proceed?

XMP: Test PF4 working with PF5 in the loop

• The successful completion of this procedure will have:

1) Demonstrated use of the PF4 coil during a plasma discharge using the “gap-control” algorithm. In this case, PF5 will be used to control the outer gap, while the PF4 is pre-programmed.

2) Demonstrated that rtEFIT correctly calculates the plasma equilibrium when PF4 is energized.

3) Demonstrated use of PF4 in isoflux control. This will verify that the line segments and process by which voltage requests are generated are working correctly.

XMP: Test PF4 working with PF5 in the loop

• Time request: ½ day (10-12 shots)

• PF-4 should be configured to be in the “pulling” direction, i.e. in opposite direction to the vertical field from PF-5 (and anti-parallel to IP). This is the standard direction for PF-4, which is usually configured to run parallel to Ip during MSE calibration.

• 1: Gap control test.

• 1.1: Load Helium gap control shot 129414 (or 132855). Verify that shot runs through. Check that rtEFIT is running and properly calculating the equilibria.

• 1.2: At start of flat-top (t=0.15 sec.), add a ramp of the PF-4 coil from 0 kA to 0.5 kA over 250 msec. This should have minimal impact on the plasma, and is a test of the ability to power the coil from within the pcc algorithm.

• 1.3: At start of flat-top, add a ramp of the PF-4 coil from 0 kA to 3 kA over 250msec. This should be a major change to the equilibrium, Test that rtEFIT is indeed calculating the equilibrium correctly. Overlay the boundaries from EFITRT, EFIT01 & 02.

• 2: Isoflux test.

• 2.1. Reload and run a standard 4MW high-delta, high-kappa morning fiducial discharge. Reduce Ip to 700 kA.

• 2.2: Add a new segment, starting on the plate at Z=80, R=140 to control the plasma boundary with PF4 coil. Turn on the PF4 control with low Proportional only (~100) gain. Test that the Isolfux algorithm can control the PF4 coil. Ramp up/down the squareness request by 0.05 during the shot to see that isoflux can follow changes in the request. Then increase by 0.1 for the final test.

XMP Feed Forward Results: PF4 scans

• PF4 between 0 and 5 kA.

• PF5 decreases as PF4 increases.

• Note that most of these are uncontrolled/unintended scans.

XMP Control Results: PF4 Initial Squareness Control

• Change squareness request while PF4 control on with new segment.

• Good control with no bias – PF3 on the other hand has bias due to controlling more than just squareness (and

not having integral gain)

• Below is a test where we changed the squareness wildly and PF4 responded well.

XMP Control Results: PF3-PF4 interaction

• With PF4 control on, we reduced the gain for PF3 %30 at 360 ms.

• PF4 compensated for the loss of inward pushing effect of PF3.– PF4 can offset both PF3 and PF5.

XMP Control Results: PF3-PF4 interaction

• Figure show the result of a ramp on PF4 from 0 to 2.6 kA.

• As PF4 increases, squareness change.

• In order to align, PF3/4/5 control points (shown in dashed black, dashed red and blue) X-point moves down.

• To solve this problem, move the PF3 and PF4 control segment. Shown in solid red, black.

Without X-point Without X-point

XP: Experimental Closed Loop System ID

• This year: Auto-tuning with Relay Feedback Method

• 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

– Relay Feedback is almost implemented on PCS.

Control Output

ProcessOutput

1. Perform relay-feedback for y1-u1 while loop 2 is on manual (Figure A)

2. Design the PI/D for u1 for based on on Kcu and Pu.

3. Perform relay-feedback for y2-u2 while loop 1 is on automatic (Figure B)

4. Design PI/D for u2.

5. Perform relay-feedback for y1-u1 while loop 2 is on automatic (Figure C)

6. Redesign PI/D for u1.

Sequential SISO

Experimental Plan for Squareness Controller

• Time request: 1day

• Load 133964 the high kappa/performance shot. Load control part of 139251. See if the shot is still the same (2 shot)

• Relay Feedback Test (1-2 shots)– This is already tested in X-point control XP.

– Start with a h value of ~200 Volts. If this is not appropriate scan h.

– Set the hysteresis value to 2*RMS measurement ~0.3/4 mWebers/rad. Test.

– Run relay-feedback on OSP with PF2L. Compare the results with already running control for OSP with PF2L (sanity check).

– Start with a small P only control for PF4 (based on the found Kcu and Pu). Test the controller is behaving as expected (correct sign and relative magnitude).

Experimental Plan for X-point Height/SP controller

• Sequential PID Tuning (8 shots)– Set PID based on Kcu and Pu. Manually tune for stability and performance.

– Relay-feedback on PF4 while PF5 control is on.

– Set PID for PF4. Manually tune for stability and performance.

– Relay-feedback for PF5 to outer gap while PF4 control is on.

– If needed repeat this process for PF4 again.

• Use PF4 to control squareness (and PF3 for more kappa) (4 shots):– Move PF3 segments 10 cm inwards (leave more room for PF4 to control the

squareness).

– Depending on the effect of PF3, detune this controller ~25%.

• Scan Squareness in the range from 0.15 to 0.5 (10 shots). – Assuming positive and negative PF4 current.

– [0.15, 0.25, 0.32, 0.37, 0.41, 0.50]

PF4 Direction?

Negative PF4 Positive PF4

• Positive – Pro: Dimple Effect Assessment: Does it have the supposed instability effect?– Con: Not much change in squareness (Shape change is due mostly to kappa)– Con: Kappa has to decrease to compensate

• Negative:– Pro: Much higher squareness range– Pro: No change in Kappa– Con: XMP was in the positive direction.

Experimental Plan for X-point Height/SP controller

• Decision Point: If the control works without problems and time permits:

Use the controller for the long pulse shots.

• X-point drifts in long pulse shots.

• We expect the results from control XMP-66 (including better axisymmetric control) and X-point/OSP

control will enhance performance

• Shot list (6-8 shots).– Load shot 135445

– Implement the XMP-66 improvements in PCS.

– Add the X-point/OSP controller for this shot

– Scan X-point from 142 to 152 along

with OSP [142, 145, 148, 152]

– Choose the best, add upper/lower X/OSP control

• Design (System ID)/Implementation of new controllers and analysis of better controlled shots.

• Eight PF coils are in the control loop (including four strike point control).

• Analysis: controlled shots show better properties than the fiducial shot (same day). • Higher confinement time than the

same day fiducial shots.

• Better response to disturbances.

• Using NSTX Toksys Model, quantify the effect of control: Decreased exponents of the unstable modes.

• Expected Results: X-point/PF4-Squareness control.

All PF coils taking part in control.

Effect of PF4 with Same Coil Currents

• Design (System ID)/Implementation of new controllers and analysis of better controlled shots.

• Eight PF coils are in the control loop (including four strike point control).

• Analysis: controlled shots show better properties than the fiducial shot (same day). • Higher confinement time than the same day

fiducial shots.

• Better response to disturbances.

• Using NSTX Toksys Model, quantify the effect of control: Decreased exponents of the unstable modes.

• Expected Results: X-point/PF4-Squareness control.

All PF coils taking part in control.

Results of NSTX Control Experiments

Current System ID

• 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

P

Kp Ki Kd

P P/Cp)(T/L) - -

PI P/Cp)(T/L) P/Cp)(3.3T/L2) -

PID P/Cp)(T/L) P/Cp)(2T/L2) P/Cp)(T/2)