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

TL;DR

This paper introduces a multi-timescale Bayesian optimization method that combines high-frequency data-driven models with low-frequency Gaussian processes to improve plasma stabilization in tokamaks, demonstrating significant performance gains.

Contribution

The paper presents a novel multi-scale Bayesian optimization framework that adaptively integrates different frequency models for effective plasma control in nuclear fusion experiments.

Findings

Outperforms baseline methods in offline tests on historical data.

Achieves a 50% success rate in live tokamak experiments.

Improves success rate by 117% over previous results.

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.