f.g. rimini – sl training 2010 – ppccpage 1/ 82 plasma position and current control f.g. rimini...

F.G. Rimini – SL Training 2010 – PPCC page 1/ 82

Plasma Position and Current Control

F.G. RiminiSL training 2010


JET plasmas can be run in a large variety of equilibria

(see P.J. Lomas’s lectures )

+

large number ( 9 + 1 ) of independently controllable circuits

Need to control in real time

• coil currents

• plasma equilibrium parameters:

Plasma current Ip

Distance of Last Closed Flux Surface (LCFS) from Plasma Facing Components (PFCs)

X-point position

Strike points positions

Why Plasma Position and Current Control:few (obvious) statements


Shape Controller (SC)

+

Vertical Stabilisation (VS)

=

PPCC

PPCC

PF coils

plasma

Magnetics


Shape Controller


in the presentation I’ll follow the thread of the SC real-time control algorithm:

• Measurement elaboration

• Control mode selection

• ( eXtreme Shape Controller module )

• Diagnostics & limit handling


Shape ControllerMeasurements elaboration


15 Internal Discrete Coils

5 Poloidal Limiter Coils

12 Ex-Vessel Saddle Loops

4 Full Flux Loops

divertor pick-up coils

x 2 octants

transmitted to SC via ATM

¡ no in-pulse check of measurements validity apart from Ip comparison !

http://users.jet.efda.org/pages/mags/ppcc/ppcc-sc/ppcc-sc.html

(KC1D) magnetic measurements for SC


SC uses pick-up (Bθ) and saddle coils (Bn) for direct derivation of

some controllable plasma paramaters:

plasma current: Ip

vertical position of current centroid: Zp( 1st order current vertical moment )

radial position of current centroid: Rp( 2nd order current vertical moment )

flux difference between two given radial positions

see extra slides for formulas

SC basic controllable variables


SC can control PF currents and/or distances between plasma boundary and wall ( = gaps )

SC - gaps determination

•Last Closed Flux Surface (LCFS) is computed by real-time boundary reconstruction code XLOC-FELIX

•XLOC approximates flux function (=solution to vacuum G-S equation) via 6th order Taylor expansion around Rp,Zp in 5 regions

•locates X-point & discriminates limiter vs X-point configuration

•computes gaps = LCFS-wall distance along pre-defined lines

•No infos on core plasma properties

! Valid for Ip ≥ 250 kA flux control used at plasma formation


Shape Controllercontroller,

control Modes

and

their selection


SC actuators (PF coils): recap

SC has control over 9 distinct circuits:

however, plasma shape is affected by all circuits real No. of degrees of freedom < 9

D1

D4

D2/D3

P1 (external)

P4 (vertical field)

P2/P3 Shaping (SHP)

P4 Imbalance ( ~radial field)

PFX ( =P1 central pancakes)


There is a strong inductive coupling between PF circuits, e.g.

P4 with PFX

SHP with PFX

P1End with PFX

D2/D3 with Imbalance

also some controllable quantities need action from several circuits:

X-point position D2 & D3

SC is a Multi Input Multi Output (MIMO) controller

SC: the controller


The output of SC is the required voltage VC for the 9 circuits

The controller is designed on the basis of :

• decoupling algorith

•(model)/measurement of gaps response to currents ( matrix B )

• measurement of mutual inductance between circuits ( matrix Ms )

• resistive compensation of the voltage drop in circuits ( matrix D )

• typical time response of circuits/gaps ( matrix C )

The set of 9 controlled variables Y is a (suitable) selection of gaps & currents IC (one is almost always Ip)

SC controller

d/dt Mff

K=MsT-1C amplifiers

Iref

IC

D

++

+

Yref

Y

+

-

see

extra

slid

es fo

r

mor

e de

tails

on

SC

cont

rolle

r


What controls what ?

Allowed combinations of actuators & controlled variables

Circuit commonly used params. / gaps

Other possible gaps

P1 IP1, Ip Primary flux

P4 flux, ROG RIG

IMB ZP, Tog Zup

PFX Current control RIG

SHP Current control UEL

D1 Current control FLX (?)

D2 & D3 Combination of ZSI (RSI) & RSO(ZSO)

Rx/Zx

D4 Current control LOG, FXO (?)

Note : gaps are in [m]


Sometimes it is necessary to use less than 9 circuits

(e.g. before / during breakdown )

In these cases, the circuits can be

Blocked = max. negative voltage ≈ open circuit

Freewheeling = zero voltage ≈ short circuit

To control or not to control ?

P4 Imbalance

activeblocked V=0

Imb PropImb Abs TogZp

Gap controlcurrent control


Control mode selection

Pulse schedule main page


Control mode selection in Pulse Schedule

SC control windows (horizontal view)

9 controllable circuitsTime

windows

Control mode choice for each time

window

Scenarios


Control mode selection in Pulse Schedule: scenarios

Scenario: pre-defined & approved set of control variables & values.

Level1 builds waveforms from values + times + transitions with previous/next time window

Values can be

changed within

limits set in the

approval document


eXtreme Shape Controller

(XSC)


Influence of current & pressure profile on shape

SC references (gaps/currents) not changed but plasma shape responds to

li (current profile)

p (pressure)


eXtreme Shape Controller (XSC)

XSC is a model-based controller developed to maintain desired plasma shape against beta, li and Ip changes(and also to test ITER relevant plasma magnetic modelling techniques)

XSC in brief:

•plasma shape in described by a set of 32 gaps / strike points / X-point position

•Singular Value Decomposition is used to identify the principal directions of the algebraic mapping between 8 coil currents and geometrical descriptors

•( actually XSC normally uses only 5-6 directions to allow for limits in the actuators)

•Control algorithm finds a compromise between actuators effort and tracking error on plasma shape

•controller works by minimizing difference between measured and desired shape


XSC integration in SC

XSC is imbedded in SC as a dedicated General Application Module (GAM)

XSC produces the current demands SC is the used in current control mode, generating the voltage requests to the amplifiers

K=MsT-1C amplifiers

IC

D

+

+

Y

+

-

XSCIref

to obtain optimal performance different controller gains need to be computed for each plasma equilibrium this XSC scenario design work is done by the SL and a PPCC expert

see extra slide


XSC selection in Pulse Schedule

XSC is activated from a Scenario Window


XSC window in Pulse Schedule

Specific XSC scenario

Gap & currents values

Allowed params.

variations

Predicted current range

Requested boundary (can be

compared with Xloc

for pulse/time)


XSC selection in Pulse Schedule

Activation of XSC for a given time window is marked by a pink box


Shape Controller

management of transitions


SC transitions = change set of control variables

Start of window Transition time:

choose wisely !


SC transitions

At the transition time the requested waveforms can be far from measured gaps/currents smoothing required

max. rate of change is specified in PPCC Expert parameters for each control variable

At transitions all waveforms are smoothed on same timescale ( based on the slowest & farthest from reference amongst the set )


Shape Controller

plasma breakdown parameters


Parameters determining the control before & during breakdown are in the main & PPCC breakdown pages of the Pulse Schedule

SC : plasma breakdown parameterswav

efor

ms

Settings of Thyristor

Make Switches (TMS) to

control dIp/dt

Premagnetisation current – can also

be selected on main page –

Breakdown mode on main page


Shape Controller

exceptions, limits …


SC limits / exceptions managementSC can encounter several types of “internal” exceptions or limits, e.g.:

•Controlled variable too far from request

•a Power Supply (PS) reaches a Voltage limit

• a PS ( not PFGC) reaches a Current limit

• P1 reaches a current limit

• Ip measurements disagree

• Magnetics data failure ( transmission )

• Failed breakdown

• Control Error

Response is adapted to severity of exception( with respect to plasma control & machine safety )


SC : classes of stops (I)

Fast Stop Triggered by PTN ( disruption imminent )

Terminate plasma with Fast Stop strategy, reduce forces at disruption ( reduce quickly Ip & elongation )

SC can :

• Issue stops to Pulse Termination Network (PTN)

• Respond to stops issued by PTN or Central Intelock and Safety System ( CISS)

Warning temporary glitch from which SC can recover

Displays message but does nothing

Slow stop Inability to perform requested control or PTN request but plasma can be terminated safely

Terminate plasma with Slow Stop strategy and send Slow Stop to PTN. Controlled termination with Ip decreased slowly and shape maintained. Changes for ILW ?


SC : classes of stops (II)

CISS-ES Software failure, cubicle power failure or externally triggered

Set all demands to zero and stops plasma immediately

Blind stop Failure of KCD1 or XLOC

Terminate plasma as a Slow Stop but with reduced set of KC1 measurements

!! Not a Stop Failure of a measurement signal during a pulse

Rare event, not detected in real time. Loss of control and disruption very likely.

P1 Stop P1 max. Limit reached ( see later )

Set PFGC into freewheeling & maintain shape

Magnetic Stop

Failure of circuit current sensors

Terminate as a Slow Stop but use only Ip, Zp & Rp control


SC limits : PS voltage/current limits

Voltage & Current are limited for each amplifier

If either limit is reached some loss of control = plasma shape different from request

! Some current limits are of both max (95%) & min (< 5%)

SC action in case of PS current limit :

• If already in current control track limit (= try to keep current constant)

• If in gap control switch to current control the gap most closely associated with the limiting circuit and track limit

• If D2/D3 are limiting Slow Stop

• If P4 is limiting Slow Stop


SC limits : P1 current limits

P1 limits are implemented in SC to avoid stresses to central solenoid

OH

net

wor

k

I I

PF

X

PF

GC

+

-

+

-

IPFX

IP1,E

IP1,C

IP1,E

min IP1,E : avoid current flowing

in opposite directions in central vs external pancakes

max IP1 : avoid coil damage due

to expanding force at high current

! max. IP1,C > IP1,E because TF

compression couteracts expansion in central pancakes

max average : avoid coil damage in case of PFX fault

! Note: min IP1,E will change if PFX On Early Timing (POET) control is implemented

see extra slide for examples of P1 limits in practice


SC exceptions: Ip consistency

For safe control and to ensure redundancy, SC has 3 independent mesurements of Ip:

• Use full (KC1D) magnetics & in-house calculation

• analogue summation of (KC1) pick-up coils integrated signals + in-house analogue compensation for MKII & Divertor currents

• Use Ip computed in Coil Protection System (CPS)

Maximum acceptable error is 4%

If difference higher than threshold for 3 consecutive cycles issue Slow Stop


SC stops : where do we get the infos ?

All stops are recorded in the Plasma protection System (PPS) mimic

Times will indicate event

sequence


Detailed infos on PPCC systems status to be found in PPCC mimic

text will become yellow in case of “events”

Click here to see all messages related to

the specific stop


… and more of the specific PPCC “events” mimics

time eventSL to verify sequence of events

what happened first ?


… what to do (in practice) …

Blind Stop

CISS-ES

Fault in magnetics, coils or SC system

Call PPCC Expert

Warning (monitor)

PPCC Slow stop

Control failure Scenario fault – SL to investigate & correct

PPCC Fast stop Disruption imminent

if plasma driven, e.g. MHD, SL to react on scenario – if caused by VS, SL to investigate but may need to call PPCC Expert


Mythbuster : SC & machine safety

Shape Controller is NOT a protection system

however it contributes to increasing machine safety:

• appropriate response to PTN stops

• stop requests sent to PTN

• amplifiers kept away from current saturation

• consistency of control with JOIs ( e.g. P1 limits )

• verify Ip measurements validity ( KC1, KC1D, CPS )

• verify magnetics data transfer ( Blind stop )


in summary


complexity of SC control can vary

very basic : Ip & plasma position

PF currents

selected gaps / PF currents

plasma boundary

and different control modes can be used for:

• different equilibria

• different phases of the plasma pulse

The SL has the responsability for

• choosing the most appropriate set of controlled variables for each phase of the plasma pulse ( & transitions )

• setting up the waveforms defining the plasma equilibria

• ( for a scenario this has to be within the constraints of the configuration approval document )

• investigating in case of SC control faults

SC in summary


Vertical Stabilisation

the basics


elongated tokamak plasmas are vertically unstable

Destabilising forces:

•Shaping Circuit

•Iron core

Stabilising forces:

•Vessel & passive structures

•Radial Field Amplifier+

+

+

Z

+

+

+δ


Vertical Stabilisation (VS) at JET

Destabilising force proportional to displacement

Instability growth is exponential: ~ o et

Growth rate = f(elongation, wall clearance, minor radius )

In JET ~ 100 – 1400 s-1 ( most plasmas ~ 150-250 s-1 )

for a typical ~200 s-1 o ~ 1 mm becomes

• 1 cm in 11 ms

• 10 cm in 23 ms

• 1 m in 34 ms

active control of vertical stability on a fast timescale is essential


Loss of Vertical Stability

When control is lost over Vertical Stabilisation the result is a particularly violent and dangerous type of disruption:

Vertical Displacement Event (VDE)

In a VDE:

• the plasma moves vertically very rapidly ( see previous slide)

• it hits the first wall before any significant loss of its thermal or magnetic energy

• with a high risk of damage to the integrity of the machine



JET system


The Session Leader has

• a large degree of control over SC

• a lot of responsability in case of some SC fault conditions

OTOH, VS is almost transparent for the SL

which means that

• the SL has very little control over VS settings

• the majority of fault conditions are handled by a PPCC Expert

• but you still have to know how it works

VS and the Session Leader


VS components

The JET VS system is composed by :

• Radial Field Coils (P2R / P3R )

• Fast amplifier for Radial Field Coils

• Dedicated fast magnetics measurements ( 4 octants )

• Digital real-time controller

Br


VS 2009 upgrade

VS has been recently upgraded with (see extra slides )• Enhaced Radial Field Amplifier (ERFA)• New hardware and software for controller (VS5)

Main problem for JET is control in cases of large perturbations ( e.g. ELMs )

old JET Vertical Stabilisation system was at the limit of its capabilities in case of large ELMs at high plasma current

risk of loss of vertical control (VDE) & damage to machine integrity



Controller


Mythbuster : JET VS controller

VS at JET does NOT control vertical position

but it does control ( 20 kHz loop):

•a suitably designed/optimised real-time estimator of the vertical speed ( vsel=0)

•the average current in the Radial Field Coils ( IERFA=Iref)

vertical position is controlled by SC on a slower timescale ( 500 Hz – 2 ms )


JET VS velocity estimator

The controlled variable for the velocity loop is an estimator of the velocity of the most unstable mode.

Historically, the value of Ip*dZp/dt has been used, computed

by flux extrapolation from the magnetic measurements

Recently alternative observers have been tested.

With ITER-like Wall, significant filtering of some signals is expected a new observer ( Obs05 ) has been designed not to use input from the filtered coils in the upper dump-plate region.

Obs05 tested in 2009 is now default controlled variable


JET VS controller loops

The VS controller output is a voltage request to ERFA

The controller structure is relatively simple:

with Igain=vgain * kI

In standard operation

• the velocity gain vgain is determined by an adaptive mechanism, i.e. it is adjusted in real time to keep the amplifier switching frequency constant

• the velocity gain is set by the adaptive mechanism to be inversely proportional to the growth rate ( loop gain ~ const. loop stability ~ const.)

• the current gain proportionality coefficient kI is given by a pre-programmed waveform

• ( the integral term is very small )

dtIKIIvvV ERFAIERFAgainselgainERFA


VS control limitations

The main perturbations to the system are:

•ELMs ( see extra slide for zoom on an ELM )

•L-H or H-L transitions ( see extra slides )

In quiescent conditions, the upgraded VS can now control plasmas with growth rate up to ~ 1400 s-1

it is still questionable whether the VS reaction to an ELM is really correct ( at least in the first few hundreds of μs )but, even with the right reaction, large RFA current excursion


Optimisation of VS behaviour at ELMs

In the past, the strategies to deal with ELMs included:

• Switch-off VS reaction (VC=0) for 1-2 ms at the ELM

• And/or apply a RFA current bias ( because ELMs always push plasma in the same direction ) – Old amplifier had limited I2t bias could be applied only for 1-2s

After VS upgrade:

• ERFA / VS5 reaction much faster

• ERFA current capability has doubled & voltage up by 20%

• ERFA has DC capability +2.5 kA bias can be used for whole heating pulse effectively 7.5 kA available

• RF circuit re-configured to give lower inductance load

• New observer obs05 seems to be less prone to wrong behaviour

• ( Alternative controllers have been tested )


VS fault conditions

Overcurrent:Smart ERFA current limit management loss of vertical control disruption likely but not necessarily imminent not triggering immediately a stop ( may benefit from DMV for mitigation )

Power Supply issues:Trip of one or more units stop request sent to PTN

Overtemperature ( unlikely with ERFA )

Trip of one or more units stop request sent to PTN

Note : with ERFA plasma can easily survive the trip of one unit PTN to be modified to have different stop in case of trip of only one unit


VS at plasma breakdownWithout plasma, magnetic configuration is vertically stable

But VS control is active before breakdown ( typically from 35s ) for Radial Field control

Sign of current loop changes when enough plasma current is measured ( magic switch parameter, settable by PPCC expert )

How much is enough ?

~ 20 – 70 kA in mode D ~ 40 kA in mode B

At breakdown, RFA current is used to compensate for field due to current induced in the MKII divertor structure.

Proportionality parameter ( baseplate prop. ) = IERFA/IMKII

• is settable by the SL ( PPCC breakdown page )

• depends on machine status ( wall conditions )

• is saved with other PPCC breakdown parameters


In summary, what can the SL do in VS ?

but you should NOT change these parameters without advice from PPCC/POG Expert

What the SL has to monitor regularly:

•ERFA current swing at ELMs ( in case of large ELMs or high )

•ERFA temperature ( frequent ELMs )

The SL has control of

•VS start time

•ERFA Current waveform ( to be used when current bias is needed )

•Baseplate proportionality ( on PPCC breakdown page)


ELM pacing with VS


Jogging the plasma up/down (= kicks ) with VS can be used to trigger ELMs & synchronise them to the jogging frequency.

The kicks parameters: Frequency, Amplitude & Duration are:

•chosen by the SL on the basis of the physics / scenario requirements in agreement/after consultation the PPCC Expert

•edited & saved by the PPCC Expert in the PPCC schedule

Use of VS for ELM pacing


the role of

the PPCC Expert


The PPCC experts are part of Plasma Operations Group (POG)

(but not everybody in POG is a PPCC Expert )

Please, do NOT think of them just as a resource to be called in case of faults !

If you think you need advice or something more than the standard SC or VS settings:

• Choice of stop strategy (SC)

• VS kicks

• XSC scenario design (SC)

• Operation with large ELMs (VS)

ask and, if necessary, involve the PPCC Expert in the session preparation with good timing ( i.e. not at the last minute )

The PPCC Expert


more infos in the practical sessions


extra slides:

formulas

gaps definitions


XLOC reconstruction is only valid once Ip ≥ few hundreds kA

alternative method to control plasma boundary is needed for early plasma phase : Flux Control

SC basic controllable variables: flux

•Measure flux at some point of the machine ( ψ1 , ψ2 )

•use extrapolation and flux difference to construct error signal to use in radial position control

drBRout 11 2

1181111 )(21

2 rBBgR

drBRin 22 2

2109922 )(21

2 rBBgR

RIF / ROF control: SC maintains equal the flux at positions R1 and R2

by expanding /moving plasma radially - no need for plasma to be

present ¡ exact plasma position is NOT controlled !


SC uses pick-up (Bθ) and saddle coils (Bn) for direct derivation of:

plasma current: Ip

dlBIII MKIIj

DjP0

4

1

1

SC basic controllable variables: Ip, Zp, Rp

dlBRR

RBZdsZJIZ nts

pp

ln

1

0

12

1,

15

1,

4

1,

kknsk

iipiMKIIMKII

jjDDjpp BBIzIzIZ

MKIIi j

jDjiip IIBI

15

1

4

1,,

dlBZRBRdsRJIRs

ntpp 21 2

0

22

kni k

skipiMKIIMKIIjDj

jDpp BBIRIRIR ,

15

1

12

1,

2,

4

1

2,

2

vertical position of current centroid: Zp ( 1st order current vertical moment )

radial position of current centroid: Rp ( 2nd order current vertical moment )


the main gaps


Plasma equilibrium

Electrical equation of system

X = gaps

M = inductance matrix R = resistance matrix (diagonal)

I = [ IP IC ] = currents V = [0 VC] = voltages

… linearisation ( & neglecting plasma resistance ) => plant model

Plasma current

Other currents

Gaps

SC: MIMO controller – plant model

),,,,( ironipPI IXX

VRIIXM )),,,(( ironip ldtd

CPCp

PI IM 1

),,( ironipCCCCs lf VIRIM

),,( ironipPCsP lII GIBX


The output of SC is the required voltage VC for the 9 circuits

The set of 9 controlled variables is a (suitable) selection of gaps

& currents (one is almost always Ip)

Resistive compensation : V=DI+U with D≈R

Plant model =>

BS & MS : gaps response to currents & mutual inductance

measured/modelled

SC: from plant model to controller

CCS

pPC

C

P

P M

I

I

TIIB

I

XY

matrixunity of rows

of rows

/

Select 9 quantities (gaps & currents) to control

CSC

dtd

dtd

UTMI

TY 1


Normalise to time constant : dY/dt = C ΔYC = user-defined diagonal matrix

Closed loop equation: UC=EC(Yref - Y)

E=MS T-1

( Add current feed-forward term Vff=Mff dIref/dt )

Complete control law is

SC: from plant model to controller

refffCrefC dtdIMDIYYECV )(

d/dt Mff

K=MsT-1C amplifiers

Iref

IC

D

++

+

Yref

Y

+

-


Control mode selection in Pulse Schedule:alternative view

9 co

ntro

llabl

e ci

rcui

ts

Time windows


Control mode selection in Pulse Schedule: waveforms

Control waveforms used by SC can be:

• Edited directly by the user

• Built by level 1 from Scenario values

• ( derived from distributed control model )

And can be viewed on the r.h.s of the main Pulse Schedule page


SC transitions = change set of control variables


XSC scenario design

to obtain optimal performance different controller gains need to be computed for each plasma equilibrium

this is done offline by a dedicated set of tools which:

• Take the desired equilibrium & a time evolution of beta/li/Ip

• Produce the specific linearized model

• Generate the set of gains (in format ready to be loaded in level 1)

• Give the boundary of applicability of the model in bp, li space

this work is done by the SL and a PPCC expert


SC limits : P1 current limits in practice

SC protective actions:

for max IP1,C or IP1,Ext or IP1,ave P1 plasma stop is issued

for min IP1,Ext track limit and remove PFX from set of controlled variables. Shape not maintaned ( Innerwall distances )

close to P1,C limit


VS components: ERFA

old FRFA ERFA

GTO thyristors

IGBT

max. output voltage

± 10 kV ± 12 kV( 4 units)

max. output current

± 2.5 kA ± 5 kA

max dV/dt 600 V/μs 800 V/μstime for full ± voltage excursion

≤ 200 μs ≤ 100 μs

• double current capability

• increase voltage by 20%

• DC capability

•faster response

and also

• re-configure Radial Field coils => lower inductance load ( faster response )

Increase:

• speed of amplifier in producing current in the coils ( ~ voltage )

• available stabilizing field ( = current )

• voltage + current size of disturbance that can be caught for a given


VS components: digital controller VS5

New hardware:

•multi-core CPU

•ATCA® based hardware

•increased data acquisition ( 192 input channels at 2 MHz )

•Up to 6 outputs

New software:

•platform independent

•modular structure

•( allows extensive offline testing )

•50 μs control loop

•up to 25 time windows per pulse

•up to 4 controllers per pulse

•up to 10 vertical velocity estimators per pulse


Zoom on an ELM

At the ELM crash VS is violently perturbed by (some or all of):

•Loss of pedestal current

•Energy drop

•Plasma jump

it is still questionable whether the VS reaction to an ELM is really correct ( at least in the first few hundreds of μs )

but, even with the right reaction, large ERFA current excursion


other examples of VS anomalous behaviour

Very frequent ELMs :

Fast switching of the amplifier could approach the temperature limit of some components

L-H / H-L transition:

Current control gain not high enough relatively large current excursion

Rotating n=2 MHD:

confusing VS and causing fast switching. Solved by 4 octants data acquisition.

f.g. rimini – sl training 2010 – ppccpage 1/ 82 plasma position and current control f.g. rimini...

Documents