george sips itpa, active control, 14 july 20031 real-time control ( and development of control...
TRANSCRIPT
George Sips ITPA, active control, 14 July 2003 1
Real-time Control
(and development of control systems)
at ASDEX Upgrade
George Sips
Max-Planck-Institut für Plasmaphysik, Boltzmannstrasse 2, D-85748, Garching,Germany.
George Sips ITPA, active control, 14 July 2003 2
Covers many physics areas,
and various different diagnostic input data in real time.
Areas for plasma control (development) at ASDEX Upgrade
• Control of plasma performance: (density profile, regime identification).
• Control of MHD instabilities: (NTM´s, sawteeth, ELM´s, ELM type).
• Control of q-profile, with on-line q(r) identification.
• Other control schemes including plasma protection.
George Sips ITPA, active control, 14 July 2003 3
Actuators at ASDEX Upgrade
George Sips ITPA, active control, 14 July 2003 4
Off axis heating: Peaking of density profile, but MHD unstable (NTM’s).
Use of central ICRH or ECRH: Can control the density peaking.
Control of density profile
AIM: Density profile control, also including the heating power as actuator
George Sips ITPA, active control, 14 July 2003 5
Control of High Z impurities in the core
Improved H-mode: NBI: 5 MW (2.5 MW off axis), ECRH, no sawteeth.
With ECRH: Density peaking reduced, no accumulation of Tungsten.
George Sips ITPA, active control, 14 July 2003 6
Regime identification
If you would like to know if the
discharge is in L-mode or H-mode.
Important for other control
schemes: Density feedback,
plasma protection, divertor.
Use of power threshold scaling law
is not sufficient at all.
H-mode or L-mode
George Sips ITPA, active control, 14 July 2003 7
Regime identification
Construct training data set >
3000 observations (by hand).
Identify:
H-mode or L-mode.
Improved H-mode or not.
Store for each data point 0-D
diagnostic data.
Automated search using
discriminant analysis on best
variables to use and optimimum
number of variables.
H-mode or L-mode dataset.
George Sips ITPA, active control, 14 July 2003 8
Regime identification
H-mode or L-mode.
Variables used for
identification
Ploss,q95,p,Ip,Vloop
Similar good results for
Improved H-mode or NotProbability x100%
George Sips ITPA, active control, 14 July 2003 9
3/2 NTM stabilisation
• More ECRH power, new mirror system for launch angle control.
• NTM detection algorithm using real time Te(r) data
George Sips ITPA, active control, 14 July 2003 10
Sawteeth control with ECCD
CounterECCD, 0.7 MW
#15847H-mode with 5 MW NBI
Sawteeth stabilisation (A. Mück):
• Counter ECCD, inside q=1 surface.
• Co ECCD, just outside q=1 surface.
• ECCD deposition tuned with B-field.
ECR deposition in pol.
-0.4 -0.2 0.0 0.2 0.4
4
0
6
2 ST/( S
T_n
o_E
CR)
New gyrotrons and mirror system
for 2003/2004
Co-ECCD
George Sips ITPA, active control, 14 July 2003 11
q95 = 3.6, sep.= 0.43, N=3.5
q95 = 3.7, sep.= 0.44, N=3.2
q95 = 4.4, sep.= 0.34, N=2.2
q95 = 3.3, sep.= 0.17, N=3.0
Control of ELM behaviour: plasma shape
George Sips ITPA, active control, 14 July 2003 12
x of the seperatrices at the outer midplane
0.5
0.4
0.3
0.2
0.1
<>
-0.06 -0.04 -0.02 0.00
x
0.2 - 0.4 ne/nGW
0.4 - 0.6 ne/nGW
0.6 - 0.8 ne/nGW
> 0.8 ne/nGW
• q95 > 3.5, pure Type II ELM´s possible.
• 0.85 < ne/nGW < 0.95.
• Close to double null shape. Type II
Control of ELM type: Type II ELM´s access conditions
George Sips ITPA, active control, 14 July 2003 13
Controlled external ELM triggering
Small pellets at 20 Hz are used to trigger ELM´s (P. Lang)
Pellet trigged ELM is similar to natural ELM.
IF E can be maintained, WELM/Wmhd can be reduced when fpellet > fnatural-ELM
A. HerrmannPSI 2002 ?
George Sips ITPA, active control, 14 July 2003 14
2 NBI sources (#6, #7) at NI-2 are off-axis,
tangential injection at 93kV, Deuterium Higher current drive efficiency With other NB sources: Control of power/particle deposition
NBI system at ASDEX Upgrade :
Motivation for current profile control at ASDEX Upgrade
George Sips ITPA, active control, 14 July 2003 15
NBI - power, source MIMO system :
)(rj
MSE, Magnetic Probe Measurements
ControllerAdaptive
PID
Statistical Methods
j (r), q (r)
magnetic pitch angle data
NBI
FP
ECCD
Current profile control at ASDEX Upgrade
George Sips ITPA, active control, 14 July 2003 16
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.5
1.0
1.5
NBI5 NBI6 NBI7 NBI8
Cu
rren
t D
ensi
ty (
MA
/m2 )
tor
total j j
NB
jBSK
0.0 0.2 0.4 0.6 0.8 1.00
2
4
6
8
10
12
Saf
ety
Fac
tor
q
tor
NBI5 NBI6 NBI7 NBI8
0.0 0.2 0.4 0.6 0.8 1.00
1
2
3
4
Ele
ctro
n T
emp
erat
ure
(ke
V)
tor
NBI5 NBI6 NBI7 NBI8
0.0 0.2 0.4 0.6 0.8 1.00
1
2
3
4
Ion
Tem
per
atu
re (
keV
)
tor
NBI5 NBI6 NBI7 NBI8
ASTRA Simulations of 4 different NBI sources
George Sips ITPA, active control, 14 July 2003 17
#17530
MSE signal : exp .VS. ASTRA
1 2 3 4 5 6 70.0
0.1
0.2
0.3
0.4S
tore
d E
ner
gy
(M
W)
Time (sec)
Exp. ASTRA with j
NB
ASTRA w/o jNB
0
5
10
15
20
PNB
(MW)
3 (MSE)
8 6 5 7 8
1 2 3 4 5 6 70
2
4
6
8
MS
E A
ng
le (
deg
)
Time (sec)
1 2 3 4 5 6 72
3
4
5
6
Channel 6 with jNB
Channel 6 w/o jNB
Channel 5 with jNB
Channel 5 w/o jNB
MS
E A
ng
le (
deg
)
Time (sec)
Astra Simulation vs Experiment at 400 kA
Different NBI sources:
George Sips ITPA, active control, 14 July 2003 18
2 4 6 8 10
Neutral beam power modulated
for optimum response on current
diffusion time scale(MW) NBIP
Time (s)
Modulated NBI
5
10
0.0 0.2 0.4 0.6 0.8 1.00.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
Cu
rren
t D
ensi
ty (
MA
/ m
2 )
r / a
ASDEX Upgrade
0.0 0.2 0.4 0.6 0.8 1.00.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
Cu
rren
t D
ensi
ty (
MA
/ m
2 )
r / a
ASDEX Upgrade
ASTRA Simulations of modulated NBI sources
These simulations are used to identify the system response function
George Sips ITPA, active control, 14 July 2003 19
7 8 9 100
5
10
PN
BI
(MW
)
Time (sec)
0.0 0.1 0.2 0.3 0.4 0.5 0.620
40
60
80
100
Test
Training
NBI5 NBI6 NBI7 NBI8
Acc
ura
cy (
%)
Minor Radius
pol
Source6 - poloidal beta
98.1 %
55.6 %
Source8 - j(0.225)
best
result
worst
result
System identification and validation with step response
George Sips ITPA, active control, 14 July 2003 20
Otrher control schemes
• Control of several neutral particle flux densities via D gas injection.
• Choice of several different line densities control via D gas injection.
(including edge density and possibility to switch during the pulse).
• Control of these line densites with pellet injection.
• Control of Tdiv (thermoelectric currents), or control of Pdiv via
impurity gas injection.
• Control of isotope mix H/(H+D).
• Reduction of Halo currents.
• Pulse repair from Deep Detachment.
George Sips ITPA, active control, 14 July 2003 21
Control of H/(H+D) mixure and line average density
George Sips ITPA, active control, 14 July 2003 22
O: Without Killer pellet.: With mode-lock detector.X: With Killer pellet.
Killer pellet activated with Neural net, trained to predict time to disruption.
But network ages rapidly, to give < 80% reliability.
Reduction of Halo current during disruptions
George Sips ITPA, active control, 14 July 2003 23
Deep detachment repair vs. Termination with killer pellet
Repair,
But experiment
compromised ?
Termination,
Experiment finished
Machine safe.
George Sips ITPA, active control, 14 July 2003 24
Design of new digital control system (ready 2003)
Termination,
Experiment finished
Machine safe.
Present system (own design): Transputer network, ~ 2 - 5 ms
New system (commercial): Shared memory netw. ~ 1 ms
George Sips ITPA, active control, 14 July 2003 25
Conclusions
Major new real time control schemes are being developed for ASDEX
Upgrade: Density profile control, Regime identification, NTM and sawtooth
stabilisation, ELM control and current profile control.
Together with existing control scheme and real time diagnostic
measurements form a consistent set of control tools.
New real time control hardware is being installed and should be ready for
commissioning in 2003