1. Page 1 ITPA IOS Kyoto, 18-21 Oct 2011 A new approach to plasma profile control in ITER S.H. Kim 1 and J.B. Lister 2 1 ITER Organization, St Paul lez Durance, France 2 EPFL-CRPP, Lausanne, Switzerland Acknowledgement : EURATOM/CRPP/CEA, FNSRS and ITER/Monaco J.-F. Artaud, V. Basiuk and F. Imbeaux (CEA) – CRONOS collaboration J.-M. Moret and O. Sauter (CRPP) – discussions D. Campbell, T.A. Casper, V. Chuyanov (ITER) – support for continuing this work

2. Page 2 ITPA IOS Kyoto, 18-21 Oct 2011 Current state of research 1.Experiments on several devices (JET, Tore-Supra, DIII-D and etc.)  Demonstrated for some cases with limited conditions 2.Two time-scale model-based profile control technique ( D. Moreau, NF 48 ) for real-time control in JET uses a profile response model deduced from the identification experiment  Concern on the range of applicability due to the evolution of the plasma state 3.Using simplified transport and source models to forecast plasma profile responses to actuator power changes in real-time ( F. Felici, NF 51, etc )  Currently in an initial phase, such as an real-time identification of transport. Not yet fully developed/tested for active profile control

3. Page 3 ITPA IOS Kyoto, 18-21 Oct 2011 Active plasma profile control in ITER Goal : robust real-time active control of multiple plasma profiles using multiple actuators 1.Robust against the evolution of the plasma state 2.Fast and simple control algorithm for real-time application 3.Handling non-linearly coupled plasma profile evolutions 4.Handling saturation and quantization of actuator powers 5.Without requiring expensive experiments and simulations Our new approach : Real-time update of Static plasma profile response models developed by Simplifying the related physics with the dual assumption of Linearity and Time-Invariance.

4. Page 4 ITPA IOS Kyoto, 18-21 Oct 2011 Te profile response model 1. The evolution of the electron pressure 2. Assuming (1) a stationary state, (2) no electron particle flux and (3) zero electron heat convective speed 4. T e profile response to the auxiliary power changes is 3. Furthermore, the auxiliary electron heat source is assumed as a product of the time-varying power and normalized radial profile shape.

5. Page 5 ITPA IOS Kyoto, 18-21 Oct 2011 q profile response model 1. 1D averaged q and plasma current density profiles 2. Directly relating the two equations and assuming the non-time-varying bootstrap current density and no radial current diffusion 3. Furthermore, assuming that the ohmic current counteracts each driven current in such a way of resulting in no edge q variation 4. q profile response to the auxiliary power changes is Controlling J pl (2 nd deriv. of psi)  target q (1 st deriv. of psi)

6. Page 6 ITPA IOS Kyoto, 18-21 Oct 2011 Required actuator power changes 1. Combining the T e and q profile responses 2. Using matrix vector notation and multiplying by weights (w e =1.0 & w q =1.0e5) and proportional gains (g e =1.0 & g q =0.5) of the control loops 3.The inverse matrix can be obtained using the SVD technique 4. The saturation of the actuator powers is taken into account by modifying the SVD calculation

7. Page 7 ITPA IOS Kyoto, 18-21 Oct 2011 Simulation of ITER hybrid mode operation 1.CRONOS (J.-F. Artaud, NF 50) used to test the new active control approach, re-using the simulation setting from DINA-CH/CRONOS hybrid mode simulations ( S.H. Kim, PPCF 51 ) 2.12MA flat-top current with 33MW of NBI and 20MW of ICRH 3.A global transport model based on the energy confinement time scaling laws, KIAUTO, is used with an assumption of improved energy confinement (H 98 =1.2) 4.Slightly reversed or flat q profiles above 1.0 at SOF

8. Page 8 ITPA IOS Kyoto, 18-21 Oct 2011 Validation of the profile response models 1.Te profile response is over-estimated in the model, due to the lack of consideration on the confinement degradation and electron-ion equipartition heating power change 2.Radial plasma current diffusion and fast evolution of the bootstrap current are not considered However, the models provide enough information for feedback control Additional 20MW EC (near on-axis) and 20MW of LH (far off-axis)

9. Page 9 ITPA IOS Kyoto, 18-21 Oct 2011 Active control of Te profile 1.Actuators : EC, IC, LH & NB 2.Slow control (> 50s) with a control interval of 10s 3.EC power is not saturated  further control of Te profile

10. Page 10 ITPA IOS Kyoto, 18-21 Oct 2011 Active control of q profile 1.Actuators : EC, LH & NB 2.Slow control (>100s) with a control interval of 10s 3.All H&CD powers are saturated at around 900s  deviation from the target

11. Page 11 ITPA IOS Kyoto, 18-21 Oct 2011 Active control of Te and q profiles T e control starts later (@ 400s) than q control (@ 300s) to avoid a strong conflict between the two controls The profiles were really ‘actively’ controlled

12. Page 12 ITPA IOS Kyoto, 18-21 Oct 2011 Are these test simulations enough? Q. Any validation experiments? A. Not yet. We would like to draw attention on this approach from experimentalists and control experts. Q. What happens if real plasma responses are different with those shown in the simulations using a global transport model? A. Let’s do another test using a physics based local transport model, GLF23

13. Page 13 ITPA IOS Kyoto, 18-21 Oct 2011 Profile response to LH application (GLF23) 1.The GLF23 transport model (0.2< ρ tor <0.95) has been used for CRONOS simulations and profile responses with 20MW of additional EC or LH are studied 2.Application of 20MW EC : similar but lower temperature profile response than the modelled one 3.Application of 20MWLH : off-axis source currents degrade the energy confinement through the evolution of magnetic shear profile ( J. Citrin, NF 50 )

14. Page 14 ITPA IOS Kyoto, 18-21 Oct 2011 Are these models not valid anymore? Q. De we need to improve the electron temperature profile response model ? A.The energy confinement dependence on the evolution of the magnetic shear profile is not yet confirmed by experiments. Real plasma profile response might be better. Therefore, this simulation can be regarded as a pessimistic (and realistic) case. Q. What happens if the partially opposite profile response is real? A. Let’s do test simulations without any modifications.

15. Page 15 ITPA IOS Kyoto, 18-21 Oct 2011 Active control of Te profile (GLF23) 1.2 cases, with and without LH, are compared 2.The electron temperature profiles were controlled 3.When LH is applied, about 2 times of LH power (20MW) were consumed At 900s, P tot with LHCD ~ 55MW P tot w/o LHCD ~ 12MW

16. Page 16 ITPA IOS Kyoto, 18-21 Oct 2011 Active control of q profile (GLF23) 1.LHCD was indispensable for controlling q profile 2.Central q value was initially increased due to the EC power requested to minimize the errors on q values outside ρ tor = 0.2 At 900s, P tot with LHCD ~ 88MW P tot w/o LHCD ~ 53MW 2 2

17. Page 17 ITPA IOS Kyoto, 18-21 Oct 2011 Active control of Te and q profiles (GLF23) T e control starts later (@ 400s) than q control (@ 300s) The profiles were controlled even in the presence of partially opposite profile response to the models

18. Page 18 ITPA IOS Kyoto, 18-21 Oct 2011 Summary and Conclusions 1.A new, simple, fast and robust approach to the real-time active control of plasma profile for ITER has been studied 2.The profiles were controllable even in the presence of partially opposite profile responses to the models. 3.A more sophisticated but less robust control technique, such as the model-based technique, can be coupled to complement each other. 4.This approach would be useful to support recently initiated ITER PCS project and IPC working group activity.

19. Page 19 ITPA IOS Kyoto, 18-21 Oct 2011 Additional slides

20. Page 20 ITPA IOS Kyoto, 18-21 Oct 2011 Auxiliary power changes (control of Te & q ) 1. NB power is initially reduced due to the strong control demand on the central q values 1 2 3 2. Actuator powers (EC, LH, NB) are quickly saturated at their upper limits 3. IC power is still available for the control of T e

21. Page 21 ITPA IOS Kyoto, 18-21 Oct 2011 Active control of Te profile (flat Xe) 1.Flat X e with a constant value has been used in evaluating the T e profile response model 2.The T e profile response model is not sensitive to the estimation of X e profile 3.Lower X e (higher C e,index (ρ)) assumption results in a slower control

22. Page 22 ITPA IOS Kyoto, 18-21 Oct 2011 NBI power quantization 1.Discrete NBI power change only when the control demand is larger than a quantized power. 2.The quantized NBI power was either 4.125MW or 8.25MW 3.A larger quantized power results in a slower control

23. Page 23 ITPA IOS Kyoto, 18-21 Oct 2011 Profile control with NBI power quantization Both Te and q profiles were well-controlled with the NBI power quantization

24. Page 24 ITPA IOS Kyoto, 18-21 Oct 2011 L-H mode transition (additional slide) 1.Feedback control is switched off at about 406s, and the auxiliary powers set to prescribed minimum values  H-L transition. 2.When the plasma is in L- mode (t=466s), the feedback control is switched on with the target Te profile at H- mode  L-H transtion 3.Te profile experienced overshooting and then it was stabilized.

25. Page 25 ITPA IOS Kyoto, 18-21 Oct 2011 Control demand (GLF23) 1.Increasing EC(red) is more effective for reducing q outside rho>0.3, while deceasing NB(blue) will only reduce central q. 2.IC(green) is more effective in increasing central Te while maintaining the temperature at the other locations.

26. Page 26 ITPA IOS Kyoto, 18-21 Oct 2011 Frequently asked questions (additional slide) 1.No experiments ? Not yet. This approach is presented in this conference for the first time, aiming at drawing attention of experts on experiments. 2.No ITB formation or L-H/H-L confinement mode transitions? Not yet. Real-time active profile control techniques and modeling capability of plasma transport are not yet advanced that much. These fast transients are usually generated in a pre-programmed manner in present-day experiments by tailoring the plasma profiles. 3.Two time-scales ? No. Fast control of kinetic profiles appears not a strong requirement for achieving and maintaining the target profiles which are linked to improved confinement regimes. It appears to be better to separately handle the fast MHD control issues However, this can be complemented using a more sophisticated control technique.