Scientific Reports volume 14, Article number: 137 (2024 ) Cite this article
Magnetic fusion plasmas, which are complex systems comprising numerous interacting elements, have large uncertainties. Therefore, future fusion reactors require prediction-based advanced control systems with an adaptive system model and control estimation robust to uncertainties in the model and observations. To address this challenge, we introduced a control approach based on data assimilation (DA), which describes the system model adaptation and control estimation based on the state probability distribution. The first implementation of a DA-based control system was achieved at the Large Helical Device to control the high temperature plasma. The experimental results indicate that the control system enhanced the predictive capability using real-time observations and adjusted the electron cyclotron heating power for a target temperature. The DA-based control system provides a flexible platform for advanced control in future fusion reactors.
Magnetic confinement fusion is a promising next-generation power source. To generate fusion-based power, the fusion plasma should be heated to attain and maintain a good state of confinement. However, fusion plasma is a typical complex system in which various physical quantities interact with each other to determine the overall behavior1. Therefore, a large number of variables should be considered for controlling the fusion plasma. An adaptive system (predictive) model is required to account for the latent variables that are difficult to observe or model (e.g., wall conditions2) along with uncertain elements related to stability, abrupt termination events, and energy and particle transport. These challenges exist in common with the control of complex systems.