Abstract:This work presents a fast and robust method for optimizing the stationary radial distribution of temperature, density and parallel current density in a tokamak plasma and its application to first-principle-based modeling of the ITER hybrid scenario. A new solver is implemented in the RAPTOR transport code, enabling direct evaluation of the stationary solution to which the radial plasma profiles evolve. Coupled to a neural network emulation of the quasi-linear gyrokinetic QuaLiKiz transport model (QLKNN-hyper-10D), a first-principle-based estimate of the stationary state of the core plasma can be found at unprecedented computational speed (typically a few seconds on standard hardware). The stationary state solver is then embedded in a numerical optimization scheme, allowing the optimization of tokamak plasma scenarios in only a few minutes. The proposed method is applied to investigate the performance of ITER hybrid scenarios at different values of total plasma current, plasma density and pedestal height and for different power contributions in a heating mix consisting of electron cyclotron and neutral beam heating. Optimizing the radial distribution of electron cyclotron current drive (ECCD) deposition, the q profile is tailored to maximize the fusion gain Q, by maximizing the energy confinement predicted through the first-principles-based transport model, while satisfying q > 1, avoiding sawtooth oscillations. It is found that optimal use of ECCD in ITER hybrid scenarios is to deposit power as close to the core as possible, while maintaining sufficient off-axis current drive to keep q above 1. Upper limits for the fusion gain Q are shown to be constrained either by minimum power requirements for the separatrix power flow to maintain H-mode or by minimum current drive requirements for q profile tailoring. Finally, it is shown that the ITER hybrid scenario operating window is significantly extended by an upgrade of the electron cyclotron power to 40 MW.

Neural-Network-Based Free-Boundary Equilibrium Solver to Enable Fast Scenario Simulations

Reconstruction of Poloidal Magnetic Fluxes on EAST based on Neural Networks with Measured Signals

An adaptive Newton-based free-boundary Grad-Shafranov solver

FreeGSNKE: A Python-based dynamic free-boundary toroidal plasma equilibrium solver

Real-time equilibrium reconstruction by neural network based on HL-3 tokamak

A new flux coordinates-based solver for fixed-boundary tokamak equilibrium with toroidal flow

Reconstruction of plasma equilibrium and separatrix using convolutional physics-informed neural operator

FEQIS: A free-boundary equilibrium solver for integrated modeling of tokamak plasmas

Real-time equilibrium reconstruction by multi-task learning neural network based on HL-3 tokamak

Neural network-augmented differentiable finite element method for boundary value problems

Physics-Informed Neural Networks for Solving 2-D Magnetostatic Fields

Grad-Shafranov equilibria via data-free physics informed neural networks

Fast equilibrium reconstruction by deep learning on EAST tokamak

Extremization to Fine Tune Physics Informed Neural Networks for Solving Boundary Value Problems

Validation of the static forward Grad-Shafranov equilibrium solvers in FreeGSNKE and Fiesta using EFIT++ reconstructions from MAST-U

Rapid optimization of stationary tokamak plasmas in RAPTOR: demonstration for the ITER hybrid scenario with neural network surrogate transport model QLKNN

Data efficiency and long term prediction capabilities for neural operator surrogate models of core and edge plasma codes

ATEQ: Adaptive Toroidal Equilibrium code

Physics-Informed Neural Network Method and Application to Nuclear Reactor Calculations: A Pilot Study

Validation of NSFsim as a Grad-Shafranov Equilibrium Solver at DIII-D

Machine learning-enhanced model-based scenario optimization for DIII-D