DIII-D RESEARCH OPPORTUNITIES FORUM FOR THE 2008 EXPERIMENTAL CAMPAIGN
Login | Review | Submit | Logout | Help
Questions about this website? Contact Andrew LeBlanc Questions about ROF? Contact Chuck Greenfield
| Title | 267: Eigenmode model-based n=1 RWM feedback control | ||
| Name: | Yongkyoon In ( |
Affiliation: | FARTECH, Inc. |
| Research Area: | Model based Control | Presentation time: | Requested | Co-Author(s): | Jin Soo Kim, Dave Humphreys, Mike Walker, Eugenio Schuster, RWM physics working group |
| Description: | The goals of this proposal are to
(1) demonstrate the superior performance of the dynamic Kalman filter based on wall surface current eigenmodes, compared to the counterpart based on picture-frame wall model (2) evaluate the performance of the eigenmode model-based PD controllers, compared with the off-line predictions (3) assess the robustness of model-based controllers | ||
| Experimental Approach/Plan: | Establish high beta, high torque RWM plasmas, where ELM-noises are dominant. It is expected that ELM-induced RWMs would readily occur. When RWM cannot be obtained in high torque plasmas, lower the plasma rotation by injecting counter beams with no magnetic braking.
Once RWM is obtained in high torque plasmas, the dominant non-RWM noise will be likely to be ELMs. Thus, the performance of the eigenmode model-based Kalman filter will be tested in comparison with picture frame model-based counterpart with respect to ELM-noise discrimination quality (related to Goal 1). Simultaneously, the closed-loop performance will be evaluated with respect to RWM suppression (related to Goal 2), which may necessitate gain scans. As the RWM experiments have been plagued by a reproducibility issue, careful comparison needs to be made even when RWM is believed to be actively stabilized. Thus, a successful RWM-free shot will be followed by an open-loop testing, which will also provide the measured open-loop RWM growth rate. Since RWM open-loop growth rate is the key parameter to relate experimental observation to eigenmode model-based modeling, it will be a good indicator how appropriately the model has been established for a given growth rate. If RWM is obtained in low torque plasmas, the non-RWM noise might not be ELMs, but fishbones or tearing modes. This will require the Kalman gain changes, so a set of the Kalman filters (e.g. based on a variety of process noise covariance magnitudes) will be prepared. The rest of the procedures will be exactly the same as mentioned for high torque case. To assess the robustness of the model-based controllers (related to Goal 3), we will do beta scans. As for high torque plasmas, the beta scan will be equivalent to growth rate scan. Thus, the beta scan can directly address the robustness of the model-based controllers, as well as allow us to assess the need of gain scheduling. Meanwhile, considering that the low torque plasmas show rather weak dependency on the RWM growth rates in a wide range of beta scans beyond ideal no-wall limit, the model-based controller performance in high C_beta is not expected to be different from that in low C_beta. As a result, we will be able to evaluate the robustness of the eigenmode model-based controllers. As a note, although the eigenmode-based DIII-D/RWM model is developed without taking into account the plasma rotation so far, the performance of eigenmode-based Kalman filter appears to fit in high torque plasmas, where an ideal MHD assumption is deemed appropriate. | ||
| Background: | In recent high beta, high torque RWM experiments, a picture-frame model (modeled open-loop RWM growth rate, gamma =120 rad/s) was experimentally confirmed to be reasonable to describe the DIII-D/RWM system, in that the dynamic Kalman filter (based on picture frame wall model) was effective in discriminating the ELM-noise from RWM [Y. In et al., Phys. Plasmas 13, 062502 (2006)].
As a more advanced model, FAR-TECH's eigenmode-based DIII-D/RWM model was predicted to be superior to picture frame wall model in terms of its effectiveness and computation times. For example, the computation times of 11 eigenmode-based Kalman filter was found to be ~ 12 us. However, considering that the performance of 3 eigenmode-based Kalman filter (i.e. requiring 6 wall states) was almost the same or slightly better than that of the successful 72 picture frame wall model mentioned above, further reduction with fewer wall states is believed to be feasible. As a result, we may shorten the computation times below DIII-D/PCS cycle time (9 us as of 2007) without compromising the capability to effectively discriminate non-RWM noise from RWM. In 2007, there was a systematic checkup for vacuum shots, whose analysis provided us with the measured L/R times, as well as the time delay between PCS commands and control coils. This helped us to properly characterize the additional connecting lines between power supply and control coils. The corresponding measurements were incorporated into DIII-D/RWM model, which led to a successful vacuum shot benchmark. Based on the latest DIII-D/RWM model, several types of model-based controllers including linear-quadratic-Gaussian (LQG) controller, are being designed in the same manner studied for preliminary model-based controllers [D. Humphreys et al., Nucl Fusion 47, 943 (2007); J. Blair et al., APS-DPP (2006)]. While all the ITER controllers are expected to be model-based, no validated RWM model exists applicable to controller design. The eigenmode-based DIII-D/RWM model is one of the candidates, which needs to be validated. | ||
| Resource Requirements: | 5 co-beams and 2-counter beams, 2 gyrotrons for ECCD | ||
| Diagnostic Requirements: | -- | ||
| Analysis Requirements: | -- | ||
| Other Requirements: | Vacuum test results prior the run-campaign and simulation results need to be confirmed first | ||
Print this page