Search

International audience; A detailed parametric analysis is presented, where the recent method based on the reservoir computing paradigm, including its statistical robustness, is studied. It is observed that the prediction capabilities of the reservoir computing approach strongly depend on the random initialization of both the input and the reservoir layers. Special emphasis is put on finding the region in the hyperparameter space where the ensemble-averaged training and generalization errors together with their variance are minimized. The statistical analysis presented here is based on the projection on proper elements method.

The exploitation of magnetically confined fusion plasmas as a sustainable and clean energy source is limited by the radially outward turbulent transport. Such transport is mainly induced by microinstabilities. Next-generation fusion devices will be mainly heated by the alpha particles born from the nuclear fusion reactions. Alpha particles must be well confined in order to transfer their energy to the bulk ions. However, very little knowledge is available regarding the interaction between alpha particles and microturbulence. Thus, unexpected turbulence and transport regimes may lead to further detrimental effects on the performance of future alpha-heated devices. The study of a tokamak scenario which can mimic the experimental conditions expected in future devices is hence crucial. Numerical investigations on the impact of fast ions on the turbulent transport driven by Ion Temperature Gradient and Trapped Electron Mode instabilities in real experiments have been carried out. It is shown that a suppression of the ion-scale turbulent transport may be achieved. Alfvén Eigenmodes (AEs) destabilized by the highly energetic ions through a wave-particle interaction play an essential role in the multi-scale mechanism leading to the turbulence suppression. Deep analyses further highlight the possibility to recognize hallmarks of the ion-scale transport reduction, regardless the dominant turbulent regime., L'exploitation des plasmas de fusion magnétiquement confinés en tant que source d'énergie durable et propre est limitée par le transport turbulent radial vers l'extérieur. Ce transport est principalement induit par les micro-instabilités. Les dispositifs de fusion de la prochaine génération seront principalement chauffés par les particules alpha produites par les réactions de fusion nucléaire. Les particules alpha doivent être bien confinées afin de transférer leur énergie aux ions thermique. Cependant, on dispose de très peu de connaissances sur l'interaction entre les particules alpha et les micro-turbulences. Ainsi, des régimes de turbulence et de transport inattendus peuvent avoir des effets néfastes sur les performances des futurs réacteurs chauffés par les particules alpha. L'étude d'un scénario de tokamak qui peut imiter les conditions expérimentales attendues dans les futurs réacteurs est donc cruciale. Des études numériques sur l'impact des ions rapides sur le transport turbulent induit par le gradient de température des ions et les instabilités du mode des électrons piégés dans des expériences réelles ont été réalisées. Il est démontré qu'une suppression du transport turbulent à l'échelle des ions peut être obtenue. Les modes d'Alfvén déstabilisés par les ions hautement énergétiques, à travers une interaction onde-particule, jouent un rôle essentiel dans le mécanisme multi-échelle qui conduit à la suppression de la turbulence. Des analyses approfondies soulignent en outre la possibilité de reconnaître les caractéristiques de la réduction du transport à l'échelle des ions, quel que soit le régime turbulent dominant.

International audience

Understanding and controlling the turbulent transport developing at the ion-scale, which strongly limits the thermal confinement in tokamaks, is crucial in view of the steady-state ITER operations. As ITER will be mainly heated by the fusion-born alpha particles, the development of ITER-relevant scenarios, together with complex numerical analyses, is crucial in order to study the possible effects of the alpha particles on turbulence regimes to date not explored in detail. Therefore, the present work aims at unveiling further steps towards the complete understanding of the impact of fast ions on the microturbulence, by extending the frontiers of our knowledge towards MeV-range of fast ion energy and turbulence patterns different to the well-established ITG, by means of very demanding gyrokinetic numerical analyses., 5th Asia Pacific Conference on Plasma Physics

International audience

A detailed parametric analysis is presented, where the recent method based on the Reservoir Computing paradigm, including its statistical robustness, is studied. It is observed that the prediction capabilities of the Reservoir Computing approach strongly depend on the random initialisation of both the input and the reservoir layers. Special emphasis is put on finding the region in the hyperparameter space where the ensemble-averaged training and generalization errors together with their variance are minimized. The statistical analysis presented here is based on the Projection on Proper Elements (PoPe) method [T. Cartier-Michaud et al., Phys. Plasmas 23, 020702 (2016)].