Multi-Timescale Dynamics Model Bayesian Optimization for Plasma Stabilization in Tokamaks

2025-06-12

AI Summary

Controlling plasma tearing mode instabilities in tokamak fusion devices remains challenging due to high measurement noise, limited experimental data, hardware constraints, and long-term dynamical drift across experimental campaigns—conditions under which conventional reinforcement learning and Bayesian optimization (BO) methods struggle to balance real-time responsiveness and robustness. Method: We propose the first multi-timescale Bayesian optimization framework, integrating a high-frequency neural ordinary differential equation (Neural ODE) dynamics model with a low-frequency self-updating sparse Gaussian process (GP), enabling online error correction and dynamic adaptation across experimental cycles. The framework unifies online Bayesian inference with closed-loop experimental control. Results: On the DIII-D tokamak, it achieves a 50% suppression success rate for unstable tearing modes—117% higher than historical benchmarks. Offline evaluations demonstrate significant superiority over state-of-the-art RL and BO baselines. This work establishes a transferable paradigm for real-time intelligent control of complex physical systems.

Technology Category

• Search and Optimization: Learning to Search
• Intelligent Robots: Learning & Optimization for ROB
• Machine Learning: Online Learning & Bandits

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Energy management for devices in mobile Web and WoT environments
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems

Abstract

Machine learning algorithms often struggle to control complex real-world systems. In the case of nuclear fusion, these challenges are exacerbated, as the dynamics are notoriously complex, data is poor, hardware is subject to failures, and experiments often affect dynamics beyond the experiment's duration. Existing tools like reinforcement learning, supervised learning, and Bayesian optimization address some of these challenges but fail to provide a comprehensive solution. To overcome these limitations, we present a multi-scale Bayesian optimization approach that integrates a high-frequency data-driven dynamics model with a low-frequency Gaussian process. By updating the Gaussian process between experiments, the method rapidly adapts to new data, refining the predictions of the less reliable dynamical model. We validate our approach by controlling tearing instabilities in the DIII-D nuclear fusion plant. Offline testing on historical data shows that our method significantly outperforms several baselines. Results on live experiments on the DIII-D tokamak, conducted under high-performance plasma scenarios prone to instabilities, shows a 50% success rate, marking a 117% improvement over historical outcomes.

Problem

• Control complex plasma instabilities in tokamaks
• Improve reliability of machine learning in nuclear fusion
• Integrate multi-scale models for real-time plasma stabilization

Innovation

• Multi-scale Bayesian optimization for plasma control
• High-frequency data-driven dynamics model integration
• Low-frequency Gaussian process adaptive updating