Session BM11: Mini-Conference: Digital Twins for Fusion Research I

High Fidelity Digital Twins of Fusion Plant

Driven by an urgent need to transition away from test-based design to a “simulation first” approach to engineering, extreme scale simulation and newly emerging AI methods offer a timely opportunity to address the challenge of designing the world’s first fusion powerplants. It is now possible to perform “High Fidelity” (HiFi) simulations of complex, strongly coupled multi-physics system of systems problems such as the tokamak. The advent of the exascale offers the possibility that simulation may finally be powerful and “actionable” enough to pre-empt emergent phenomena that are typically only discovered once plant becomes operational. At scale, by combining detailed CAD models that closely resemble real plant with fully coupled multi-physics models and access to ExaFlop GPU supercomputers, models will start to closely resemble Grieves’ and Vickers’ definition of the “digital twin” [1].

Several UKAEA facilities are planning digital twins, for example CHIMERA [2], a multi-physics thermal hydraulic / electromagnetic test facility, and MAST-U [3], UKAEA’s Spherical Tokamak. These are being used as test cases to assess the utility of HiFi digital twins and their promise for accelerating and de-risking commercial fusion. A key ingredient will be emulators such as GANNs, PINNs or other AI methods (e.g. neural PDE operators) that can replace, augment or accelerate the more traditional simulation tools.

[1] Grieves, M., Vickers, J. (2017). Digital Twin: Mitigating Unpredictable, Undesirable Emergent Behavior in Complex Systems. In: Kahlen, J., Flumerfelt, S., Alves, A. (eds) Transdisciplinary Perspectives on Complex Systems. Springer, Cham. https://doi.org/10.1007/978-3-319-38756-7_4

[2] T.R. Barrett, M. Bamford, B. Chuilon, T. Deighan, P. Efthymiou, L. Fletcher, M. Gorley, T. Grant, T. Hall, D. Horsley, M. Kovari, M. Tindall, The CHIMERA facility development programme, Fusion Engineering and Design, Volume 194, 2023, 113689, ISSN 0920-3796, https://doi.org/10.1016/j.fusengdes.2023.113689.

[3] Morris, William, et al. "MAST upgrade divertor facility: a test bed for novel divertor solutions." IEEE transactions on plasma science 46.5 (2018): 1217-1226.   

Digital Twin framework development for SPARC and ARC

The mission of the SPARC tokamak is to close critical science and technology gaps needed to design and operate the ARC power plant, enabled by quickly reaching the maximum projected performance of many SPARC systems. This goal is likely to be accelerated if SPARC can deploy a digital twin that can be updated rapidly in parallel with operations, quantify the machine state with respect to operational budgets and component lifetime, and fold in new data as operational space is expanded.  Demonstrating the value of a digital twin on SPARC would also lay the groundwork for use on ARCs where digital twins can aid in constraining machine state awareness for a very sparse diagnostic set and can learn from a fleet of power plants.  To build such a framework will require unifying physics and engineering models across a wide range of spatial and temporal scales, and it will require harmonization of many different types of computational algorithms into a cohesive framework.

In this work we outline the broader context in which CFS sees the potential for digital twins to make an impact on fusion energy, and highlight the challenges that a SPARC digital twin framework could address, using one of the most challenging key issues for a next step device, power exhaust, as an example.     

A Vision for Simulation-based, Multi-fidelity Digital Twins in Fusion Energy

Fusion energy is a unique field where the combination of rich multimodal diagnostics and multi-physics, multi-fidelity simulation are vital to progress. Here we present several efforts and techniques pursued towards building out faithful digital twins, based on a range of simulations covering differing physics and levels of fidelity. These digital twins can be leveraged for eventual use in downstream tasks such as design optimization, scenario planning, experiment interpretation, and control. 

Work done through a multi-institution SciDAC project called StellFoundry is building up a framework for faithful digital models of stellarators, covering the self-consistent physics predictions from the fusion power core all the way to engineering calculations utilizing neutron and heat loads. Tools and schema for coupling these codes in HPC environments for multi-physics simulation capability are being developed, in addition to advanced optimization techniques and AI surrogates, to include higher-fidelity simulation in optimization loops. 

Digital twins can be comprehensive or focused on targeted subsystems. In another work we create an AI model for accelerated predictions of the HEAT code, used for calculations of the divertor heat load on the upcoming SPARC tokamak. For a fixed divertor design, this model can predict CAD-level, detailed divertor heat loads, useful for engineering operation monitoring, and reduction into control algorithms.  

Grounding digital models with data from the physical asset is a key requirement for digital twins, to account for gaps in physical realism of simulation models. Simulation-based inference based on AI models enables fast Bayesian inference of digital twin physics information from a combination of diagnostics, and opens a path for grounding of the models in addition to uncertainty quantification.    

Progress Towards a Predictive DIII-D Digital Twin with Neutral Beam Heat-Load and As-built Geometry

A selection of recent experiments at the DIII-D tokamak are digitally recreated in high-fidelity using as-built device data, demonstrating the impact on experimental decision making using a digital twin. The first is an investigation of a series of shots with neutral beam heating and excessively high carbon impurities, where a digital twin can avoid the use of experimental run time to determine which beam is likely creating the impurity source. The second is where heating inside a beam duct created a localized hot-spot that exceeded nominal limits. Presented here are concrete examples of how a digital twin can be used to improve operational efficiency and altogether avoid scenarios that present a high risk to the integrity of the device. To achieve this, a digital twin of DIII-D is being constructed by integrating the particle following code IonOrb with sub-millimeter accuracy laser scan data of the as-built first wall, and extending it beyond the toroidal field coils to include the ducts connecting the neutral beam injectors with the torus. Progress will be reported on the integration of a data-driven predictive equilibrium (using selective state-space models) to offer a prediction of neutral beam heat-load prior to the planned plasma discharge.   

Digital twin control for LHD plasmas based on data assimilation system ASTI

The operation of future fusion reactors requires a system that predicts and controls fusion plasma behavior under conditions of limited observation. To address this challenge, we have introduced a control approach based on data assimilation (DA), which integrates predictive model (digital twin) adaptation using real-time measurements and control estimation robust to model and observation uncertainties. The main part of the DA-based control system, ASTI, computes many integrated simulations in real-time to predict the probability distribution of future plasma states. In addition, the system estimates the optimal control input and the actual plasma state based on Bayes' theorem. The DA-based control system implemented in the Large Helical Device (LHD) was successfully applied to control the electron temperature using the electron cyclotron heating and the real-time Thomson scattering measurement. In the control experiment, it was demonstrated that the digital twin's predictive capability was improved by optimizing the model parameters using the real-time observations. The DA-based control system allows for the harmonic connection of measurement, heating, fueling, and simulation and can provide a flexible platform for digital twin control of future fusion reactors. In this talk, we will discuss the details of the control system and the demonstration experiments.   

Progresses in the Development of a Virtual Tokamak Platform based on Digital Twin Technology

Digital twin technology can be defined as a collection of IT technologies which enable a virtualization of the shapes and functions of objects in digital space. Along with the steady progresses in supercomputing technologies, recent advancements in large scale GPU computing enable a digital twining of highly complex objects comprising a large scale plant. In this presentation, we report recent progresses in the development of a virtual tokamak platform based on the digital twin technology. Previously, we have identified and developed several enabling IT technologies for fusion digital twin, and applied them to implement the Virtual KSTAR (V-KSTAR) as the digital twin of the KSTAR tokamak[1, 2]. Two different modes of operation were provided by this first version of the V-KSTAR: 1) real time monitoring of machine operation and experiment and 2) integrated 3D visualization and analysis of fusion simulation. Recently, a new capability was added as the 3^rd mode of operation 3) diagnostics simulation on KSTAR. As a demonstrative example of the 3^rd operational mode, we introduce the application of the V-KSTAR for the development of a new Lyman alpha diagnostics on KSTAR. Though the development of the virtual tokamak platform was started with the KSTAR tokamak as the first target device, our development is ultimately aiming a general platform. We also report the progresses in the generalization of the software designs toward a general platform, which is applicable for any tokamak device.   

FUSE: digital twin framework for tokamak fusion power plant design and operations

The FUsion Synthesis Engine (FUSE) integrates first-principle, machine-learning, and reduced models into a comprehensive simulation tool for physics, engineering, control, costing, and risk assessment. Originally designed for integrated FPP machine design, FUSE is now being extended to model the time evolution of the plasma and plant, supporting simulations in both feedforward and feedback modes with controllers. FUSE employs a genetic algorithm to identify viable machine designs that optimize performance, cost, and risk objectives while adhering to physics, engineering constraints, and stakeholder requirements. Comparative trade studies on positive and negative triangularity FPP designs, along with pulse trajectory optimization to maximize ITER fusion burn, will be discussed. Coupling FUSE to the TokSys control environment is under way to enable operators to test control strategies, predict outcomes, and adjust parameters in a risk-free virtual environment. Initial results will be discussed. FUSE exemplifies the potential of digital twins in fusion research, supporting decision-making, and ultimately accelerating fusion commercialization.   

Advancing tokamak research and development through tailored digital twins and Fusion Twin Software as a Service (SaaS)

Digital twins are often tailored to meet diverse stakeholder needs, from research to predictive maintenance. They offer precise insights and actionable data, improving decision-making and driving innovation across various domains.

To create digital twins of tokamaks, Next Step Fusion applies NSFsim, a flexible and fast control-oriented simulator. It solves 2D free-boundary plasma equilibrium (Grad-Shafranov equation) together with transport equations for a digital replica of a real device.

Precise simulations alone cannot address all needs. Fusion Twin SaaS was launched to bridge the gap between real machines, simulators, and digital twins. It provides cloud-based simulations of multiple tokamaks with NSFsim, plasma boundary reconstruction machine learning (ML) model, data management and visualization, Jupyter Notebooks, and collaboration features.

Future plans include creating an environment for training ML models, adding more digital replicas of real and virtual tokamaks, extending simulator capabilities with better transport and auxiliary heating and current drive models, and making available control, reconstruction, and prediction ML models. This will speed up research, reduce time to insights, simplify work for fusion professionals, and eventually bring fusion energy closer.