Real-time pedestal optimization and ELM control with 3D fields and gas flows on DIII-D

Abstract

The capabilities of the DIII-D tokamak's plasma control system (PCS) were expanded to allow for pedestal optimization and edge localized mode (ELM) control. Three proof of principle control schemes are presented that were successfully implemented and tested. These use multiple inputs from real-time (RT) diagnostics like a D α based ELM monitor and edge profile measurements from Thomson scattering (TS) as well as 3D, i.e. non-axisymmetric, magnetic fields and gas puffs as actuators to regulate the density pedestal.The first scheme targets to optimize the access of ELM suppression induced by non-axisymmetric magnetic perturbations (MPs). The conducted set of experiments identifies a path dependence of plasma confinement on the applied MP amplitude. The controller aims to transition into ELM suppression at the minimum 3D field amplitude and reduces it further afterwards, allowing for partial confinement recovery. Another pedestal control scheme is deployed to compensate the density 'pump-out' in MP ELM suppression by regulating the gas puff. This uses RT TS diagnostic data, extracting the pedestal height from the electron density (n e ) profiles and enables studies of the transition into and out of MP ELM suppression at constant density. A limit cycle behavior of edge rotation and MP amplitude persists under these conditions. The third control scheme combines MPs and gas puffs as actuators to perform pedestal density trajectory control to access Super high confinement mode (H-mode) and furthermore, allowing the integration of a radiative divertor in this regime. While MPs mainly impact the pedestal top density, the control scheme allows to loosen the tight coupling of pedestal top and separatrix density evolution.With respect to ITER, the achieved results emphasize the need for an advanced control system to keep MP amplitude close to but above the ELM suppression threshold at all times, enabling high confinement and, respectively, at high fusion energy gain factor (Q). Furthermore, pedestal control enables detailed physics studies in present-day tokamaks and allows the exploration of core-edge integrated plasma scenarios.

DIII-D research advancing the physics basis for optimizing the tokamak approach to fusion energy

“…In addition, target T e and ion saturation current (J sat ) from floor probes, and the observed 30% reduction of the measured outer divertor strike point (OSP) heat flux, confirmed the OSP was at the onset of detachment during the time the high pedestal pressure was maintained [58]. Advanced control algorithms [61,62] were used to achieve these results including the use of feedback-controlled 3D fields for density control and feedback nitrogen gas puffing for divertor radiated power control. All of these results suggest that it may be desirable to look into SH-like pedestal pressure enhancements in ITER scenarios with detached radiative divertors.…”

Preemptive RMP-driven ELM crash suppression automated by a real-time machine-learning classifier in KSTAR

“…The back-transition event via a hardware fault is dealt with by the interlock system, so the in-vessel control coils cease automatically upon reaching the end of the pulse. Note also that the method differs from the known ELM control algorithms that mainly aim to regulate the ELM frequency [27] or pedestal [28], which all assume that an H-mode pedestal with detectable ELM bursts already exists. The most significant advantage of the ML-based algorithm is to suppress preemptively before the pedestal height becomes excessive and thereby alter the characteristic properties of the pedestal during the buildup of the H-mode profiles, as we show in the next section.…”

Data-driven profile prediction for DIII-D

“…This choice is still relevant for control, as we can use total power and torque as a proxy for individual beam settings via the existing DIII-D 'VEP' controller [37]. Similarly, as a proxy for individual gas flow rates, we use the target for the existing DIII-D line-averaged density controller (see [38] and references therein for details on the density controller).…”