The TCV Team | Searcharxiv

arXiv2026

The physics of ELM-free regimes in EUROfusion tokamaks

The development of operational scenarios without large Type-I ELMs is of utmost importance for the stable operation and longevity of future tokamaks. The EUROfusion tokamak exploitation program has therefore made the understanding of ELM-free regimes a major topic of exploration across all its contributing devices (ASDEX Upgrade, JET, MAST-Upgrade, TCV, and WEST). An integrated program to investigate a range of Type-I ELM-free regimes has been developed covering the enhanced D-alpha (EDA), magnetic perturbations (MP), negative triangularity (NT), quasi-continuous exhaust (QCE), quiescent H-mode (QH), the baseline small ELMs (SE), I-mode, and X-point radiator (XPR) regimes. This contribution focuses on the development and understanding of the NT and QCE regimes on ASDEX Upgrade, JET, and TCV. The importance of transport via ballooning modes in both regimes is highlighted, as well as the progress in developing access models based on ideal-MHD. In the case of the QCE, this can also be expressed as a minimum separatrix density, which corresponds well to experimentally measured separatrix densities. Particular focus is paid to the performance of the QCE in terms of the achieved pedestal top values, which, when appropriately normalised, do not differ significantly from ELMy H-mode plasmas. This, combined with the predicted minimum separatrix density for the 15 MA ITER baseline plasma, highlight the relevance of the QCE as a potential operational scenario for both ITER and future reactors.

physics.plasm-ph↗

arXiv2023

Towards practical reinforcement learning for tokamak magnetic control

Reinforcement learning (RL) has shown promising results for real-time control systems, including the domain of plasma magnetic control. However, there are still significant drawbacks compared to traditional feedback control approaches for magnetic confinement. In this work, we address key drawbacks of the RL method; achieving higher control accuracy for desired plasma properties, reducing the steady-state error, and decreasing the required time to learn new tasks. We build on top of \cite{degrave2022magnetic}, and present algorithmic improvements to the agent architecture and training procedure. We present simulation results that show up to 65\% improvement in shape accuracy, achieve substantial reduction in the long-term bias of the plasma current, and additionally reduce the training time required to learn new tasks by a factor of 3 or more. We present new experiments using the upgraded RL-based controllers on the TCV tokamak, which validate the simulation results achieved, and point the way towards routinely achieving accurate discharges using the RL approach.

physics.plasm-ph↗