Understanding and control of electron thermal transport is a critical point of research in magnetic fusion experiments. Previous experiments have shown that operation with reversed magnetic shear (RMS) can suppress electron thermal transport, resulting in the generation of internal transport barriers (ITBs), with the location of the ITB correlated with the location of minimum magnetic shear. The recent upgrades to NSTX-increased magnetic field up to 1 T, increased plasma current up to 2 MA, 2nd neutral beam-present an increased operating space in which to explore electron thermal transport in RMS plasmas. Utilizing TRANSP, we have developed operating scenarios by which to generate RMS in NSTX-U. The results suggest that RMS in NSTX-U can be generated through fast current ramp and early beam injection into a large plasma volume. This is very similar to the procedure that was followed in both TFTR and NSTX to generate RMS. Sustainment of RMS, disregarding non-( qmin = 1) MHD events, requires maintaining a large plasma volume, and increasing the core Te, either via increased plasma current and/or adding heating power. Using this procedure, RMS was sustained for similar to 1 s, with qmin > 1 for that period.

The rampdown phase of a tokamak pulse is difficult to simulate and often exacerbates multiple plasma instabilities. To reduce the risk of disrupting operations, we leverage advances in Scientific Machine Learning (SciML) to combine physics with data-driven models, developing a neural state-space model (NSSM) that predicts plasma dynamics during Tokamak à Configuration Variable (TCV) rampdowns. The NSSM efficiently learns dynamics from a modest dataset of 311 pulses with only five pulses in a reactor-relevant high-performance regime. The NSSM is parallelized across uncertainties, and reinforcement learning (RL) is applied to design trajectories that avoid instability limits. High-performance experiments at TCV show statistically significant improvements in relevant metrics. A predict-first experiment, increasing plasma current by 20% from baseline, demonstrates the NSSM's ability to make small extrapolations. The developed approach paves the way for designing tokamak controls with robustness to considerable uncertainty and demonstrates the relevance of SciML for fusion experiments.

Typically the rampdown phase of a shot consists of a decrease in current and injected power and optionally a change in shape, but there is considerable flexibility in the rate, sequencing, and duration of these changes. On the next generation of tokamaks it is essential that this is done safely as the device could be damaged by the stored thermal and electromagnetic energy present in the plasma. This works presents a procedure for automatically choosing experimental rampdown designs to rapidly converge to an effective rampdown trajectory. This procedure uses probabilistic machine learning methods paired with acquisition functions taken from Bayesian optimization. In a set of 2022 experiments at DIII-D, the rampdown designs produced by our method maintained plasma control down to substantially lower current and energy levels than are typically observed. The actions predicted by the model significantly improved as the model was able to explore over the course of the experimental campaign.

The DIII-D tokamak has elucidated crucial physics and developed projectable solutions for ITER and fusion power plants in the key areas of core performance, boundary heat and particle transport, and integrated scenario operation, with closing the core-edge integration knowledge gap being the overarching mission. New experimental validation of high-fidelity, multi-channel, non-linear gyrokinetic turbulent transport models for ITER provides strong confidence it will achieve Q 10 operation. Experiments identify options for easing H-mode access in hydrogen, and give new insight into the isotopic dependence of transport and confinement. Analysis of 2,1 islands in unoptimized low-torque IBS demonstration discharges suggests their onset time occurs randomly in the constant beta phase, most often triggered by non-linear 3-wave coupling, thus identifying an NTM seeding mechanism to avoid. Pure deuterium SPI for disruption mitigation is shown to provide favorable slow cooling, but poor core assimilation, suggesting paths for improved SPI on ITER. At the boundary, measured neutral density and ionization source fluxes are strongly poloidally asymmetric, implying a 2D treatment is needed to model pedestal fuelling. Detailed measurements of pedestal and SOL quantities and impurity charge state radiation in detached divertors has validated edge fluid modelling and new self-consistent 'pedestal-to-divertor' integrated modeling that can be used to optimize reactors. New feedback adaptive ELM control minimizes confinement reduction, and RMP ELM suppression with sustained high core performance was obtained for the first time with the outer strike point in a W-coated, compact and unpumped small-angle slot divertor. Advances have been made in integrated operational scenarios for ITER and power plants. Wide pedestal intrinsically ELM-free QH-modes are produced with more reactor-relevant conditions, Low torque IBS with W-equivalent radiators can exhibit predator-prey oscillations in T-e and radiation which need control. High-beta(P) scenarios with q(min) > 2, q(95)-7.9, beta(N) > 4, beta(T)-3.3% and H-98y2 > 1.5 are sustained with high density ((n) over bar = 7E19 m(-3), f(G)-1) for 6 tau(E), improving confidence in steady-state tokamak reactors. Diverted NT plasmas achieve high core performance with a non-ELMing edge, offering a possible highly attractive core-edge integration solution for reactors.

The objectives of NSTX-U research are to reinforce the advantages of STs while addressing the challenges. To extend confinement physics of low- A , high beta plasmas to lower collisionality levels, understanding of the transport mechanisms that set confinement performance and pedestal profiles is being advanced through gyrokinetic simulations, reduced model development, and comparison to NSTX experiment, as well as improved simulation of RF heating. To develop stable non-inductive scenarios needed for steady-state operation, various performance-limiting modes of instability were studied, including MHD, tearing modes, and energetic particle instabilities. Predictive tools were developed, covering disruptions, runaway electrons, equilibrium reconstruction, and control tools. To develop power and particle handling techniques to optimize plasma exhaust in high performance scenarios, innovative lithium-based solutions are being developed to handle the very high heat flux levels that the increased heating power and compact geometry of NSTX-U will produce, and will be seen in future STs. Predictive capabilities accounting for plasma phenomena, like edge harmonic oscillations, ELMs, and blobs, are being tested and improved. In these ways, NSTX-U researchers are advancing the physics understanding of ST plasmas to maximize the benefit that will be gained from further NSTX-U experiments and to increase confidence in projections to future devices.

Real time detection of time evolving growth rates of multiple stable magnetohydrodynamic (MHD) eigenmodes has been achieved in DIII-D tokamak experiments via multi-mode three-dimensional (3D) active MHD spectroscopy. The measured evolution of the multi-modes’ growth rates is in good accordance with the variation of the plasma β N . Using experimental equilibria, resistive MARS-F simulations found the two least stable modes to have comparable growth rates to those experimentally measured. Real time and offline calculations of the modes’ growth rates show comparable results and indicate that cleaner system input and output signals will improve the accuracy of the real time stability detection. Moreover, the shortest real time updating time window of multi-mode eigenvalues can be about 2 ms in DIII-D experiments. This real time monitoring of stable, macroscopic kink and tearing modes thus provides an effective tool for avoidance of the most common causes of tokamak disruption.

Deep learning approaches have been applied to images of C III emission in the lower divertor of DIII-D to develop models for estimating the level of detachment and magnetic configuration (X-point location and strike point radial location). The poloidal distance from the target to the C III emission front is used to represent the level of detachment. The models perform well on a test dataset not used in training, achieving F _1 scores as high as 0.99 for detachment state classification and root mean squared error (RMSE) as low as 2 cm for front location regression. Predictions for shots with intermittent reattachment are studied, with class activation mapping used to aid in interpretation of the model predictions. Based on the success of these models, a third model was trained to predict the X-point location and strike point radial position from C III images. Though the dataset covers only a small range of possible magnetic configurations, the model shows promising results, achieving RMSE around 1 cm for the test data.

Control for tokamaks, the leading candidate technology for nuclear fusion, is an important pursuit since the realization of nuclear fusion as an energy source would result in virtually unlimited clean energy. However, control of these devices remains a challenging problem due to complex, non-linear dynamics. At the same time, there is promise in learning controllers for difficult problems thanks to recent algorithmic developments in reinforcement learning. Because every run (or shot) of the tokamak is extremely expensive, in this work, we investigated learning a controller from logged data before testing it on a tokamak. In particular, we used 18 years of data from the DIII-D device in order to learn a controller for the neutral beams that targets specified beta(N) (normalized ratio of plasma pressure to magnetic pressure) and rotation quantities. This was done by using the data to first learn a dynamics model, and then by using this model as a simulator to generate experience to train a controller via reinforcement learning. During a control session on DIII-D, we tested both the ability for our dynamics model to design feedforward trajectories and the controller's ability to do feedback control to achieve specified targets. This work marks some of the first steps in doing reinforcement learning for tokamak control through historical data alone.

The neutral beam injection (NBI) system in EAST produces energetic neutral particles, which collide with electrons and ions in tokamak plasmas and heat the plasmas through Coulomb collisions. Moreover, it drives a non-inductive source of current, due to the charge-exchange collision between neutral particles and ions, and injects toroidal torque, which generates a toroidal rotation of the plasma. The effect caused by the NBI system, such as plasma heating, current drive, total neutron rate, momentum transfer, and shine-through, are modeled by a comprehensive module called NUBEAM. However, NUBEAM is computationally intensive since it relies on Monte Carlo methods. In this work, a neural network model has been developed as a surrogate model for NUBEAM in EAST. The database for neural-network model training, validation and testing is generated by running TRANSP for experimental discharges from recent EAST campaigns (after the latest NBI upgrade) while using the NUBEAM module. Simulation results illustrate that the trained neural network has the capability of replicating the predictions made by NUBEAM while demanding a significantly shorter execution time. These results indicate that surrogate models like the one proposed in this work could enable fast transport simulations for EAST after integrating them into a control-oriented predictive code such as COTSIM.

This paper describes a real-time capable algorithm for identifying the safe operating region around a tokamak operating point. The region is defined by a convex set of linear constraints, from which the distance of a point from a disruptive boundary can be calculated. The disruptivity of points is calculated from an empirical machine learning predictor that generates the likelihood of disruption. While the likelihood generated by such empirical models can be compared to a threshold to trigger a disruption mitigation system, the safe operating region calculation enables active optimization of the operating point to maintain a safe margin from disruptive boundaries. The proposed algorithm is tested using a random forest disruption predictor fit on data from DIII-D. The safe operating region identification algorithm is applied to historical data from DIII-D showing the evolution of disruptive boundaries and the potential impact of optimization of the operating point. Real-time relevant execution times are made possible by parallelizing many of the calculation steps and implementing the algorithm on a graphics processing unit. A real-time capable algorithm for optimizing the target operating point within the identified constraints is also proposed and simulated.

DIII-D physics research addresses critical challenges for the operation of ITER and the next generation of fusion energy devices. This is done through a focus on innovations to provide solutions for high performance long pulse operation, coupled with fundamental plasma physics understanding and model validation, to drive scenario development by integrating high performance core and boundary plasmas. Substantial increases in off-axis current drive efficiency from an innovative top launch system for EC power, and in pressure broadening for Alfven eigenmode control from a co-/counter- I p steerable off-axis neutral beam, all improve the prospects for optimization of future long pulse/steady state high performance tokamak operation. Fundamental studies into the modes that drive the evolution of the pedestal pressure profile and electron vs ion heat flux validate predictive models of pedestal recovery after ELMs. Understanding the physics mechanisms of ELM control and density pumpout by 3D magnetic perturbation fields leads to confident predictions for ITER and future devices. Validated modeling of high- Z shattered pellet injection for disruption mitigation, runaway electron dissipation, and techniques for disruption prediction and avoidance including machine learning, give confidence in handling disruptivity for future devices. For the non-nuclear phase of ITER, two actuators are identified to lower the L–H threshold power in hydrogen plasmas. With this physics understanding and suite of capabilities, a high poloidal beta optimized-core scenario with an internal transport barrier that projects nearly to Q = 10 in ITER at ∼ 8 MA was coupled to a detached divertor, and a near super H-mode optimized-pedestal scenario with co- I p beam injection was coupled to a radiative divertor. The hybrid core scenario was achieved directly, without the need for anomalous current diffusion, using off-axis current drive actuators. Also, a controller to assess proximity to stability limits and regulate β N in the ITER baseline scenario, based on plasma response to probing 3D fields, was demonstrated. Finally, innovative tokamak operation using a negative triangularity shape showed many attractive features for future pilot plant operation.

The mission of the low aspect ratio spherical tokamak NSTX-U is to advance the physics basis and technical solutions required for optimizing the configuration of next-step steady-state tokamak fusion devices. NSTX-U will ultimately operate at up to 2 MA of plasma current and 1 T toroidal field on axis for 5 s, and has available up to 15 MW of neutral beam injection power at different tangency radii and 6 MW of high harmonic fast wave heating. With these capabilities NSTX-U will develop the physics understanding and control tools to ramp-up and sustain high performance fully non-inductive plasmas with large bootstrap fraction and enhanced confinement enabled via the low aspect ratio, high beta configuration. With its unique capabilities, NSTX-U research also supports ITER and other critical fusion development needs. Super-Alfvénic ions in beam-heated NSTX-U plasmas access energetic particle (EP) parameter space that is relevant for both α-heated conventional and low aspect ratio burning plasmas. NSTX-U can also generate very large target heat fluxes to test conventional and innovative plasma exhaust and plasma facing component solutions. This paper summarizes recent analysis, theory and modelling progress to advance the tokamak physics basis in the areas of macrostability and 3D fields, EP stability and fast ion transport, thermal transport and pedestal structure, boundary and plasma material interaction, RF heating, scenario optimization and real-time control.

Many potential applications of reinforcement learning (RL) are stymied by the large numbers of samples required to learn an effective policy. This is especially true when applying RL to real-world control tasks, e.g. in the sciences or robotics, where executing a policy in the environment is costly. In popular RL algorithms, agents typically explore either by adding stochasticity to a reward-maximizing policy or by attempting to gather maximal information about environment dynamics without taking the given task into account. In this work, we develop a method that allows us to plan for exploration while taking both the task and the current knowledge about the dynamics into account. The key insight to our approach is to plan an action sequence that maximizes the expected information gain about the optimal trajectory for the task at hand. We demonstrate that our method learns strong policies with 2x fewer samples than strong exploration baselines and 200x fewer samples than model free methods on a diverse set of low-to-medium dimensional control tasks in both the open-loop and closed-loop control settings.

Neural networks (NNs) offer a path towards synthesizing and interpreting data on faster timescales than traditional physics-informed computational models. In this work we develop two NNs relevant to equilibrium and shape control modeling, which are part of a suite of tools being developed for the National Spherical Torus Experiment-Upgrade for fast prediction, optimization, and visualization of plasma scenarios. The networks include Eqnet , a free-boundary equilibrium solver trained on the EFIT01 (Equilibrium FITtting 01) reconstruction algorithm, and Pertnet , which is trained on the Gspert code and predicts the non-rigid plasma response, a nonlinear term that arises in shape control modeling. The NNs are trained with different combinations of inputs and outputs in order to offer flexibility in use cases. In particular, Eqnet can use magnetic diagnostics as inputs and act as an EFIT-like reconstruction algorithm, or, by using pressure and current profile information the NN can act as a forward Grad–Shafranov equilibrium solver. This forward-mode version is envisioned to be implemented in the suite of tools for simulation of plasma scenarios. The reconstruction-mode version gives some performance improvements compared to the online reconstruction code real-time EFIT, especially when vessel eddy currents are significant. We report strong performance for all NNs indicating that the models could reliably be used within closed-loop simulations or other applications. Some limitations are discussed.

GPECnet is a densely connected neural network that has been trained on GPEC data, to predict the plasma stability, neoclassical toroidal viscosity (NTV) torque, and optimized 3D coil current distributions for desired NTV torque profiles. Using NTV torque, driven by non-axisymmetric field perturbations in a tokamak, can be vital in optimizing pedestal performance by controlling the rotation profile in both the core, to ensure tearing stability, and the edge, to avoid edge localized modes (ELMs). The generalized perturbed equilibrium code (GPEC) software package can be used to calculate the plasma stability to 3D perturbations and the NTV torque profile generated by applied 3D magnetic fields. These calculations, however, involve complex integrations over space and energy distributions, which takes time to compute. Initially, GPECnet has been trained solely on data representative of the quiescent H-mode (QH) scenario, in which neutral beams are often balanced and toroidal rotation is low across the plasma profile. This work provides the foundation for active control of the rotation shear using a combination of beams and 3D fields for robust and high performance QH mode operation.