Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Plasma Confinement State Classification in Fusion Power Plants: Profile Reflectometer and Ensemble Diagnostics

Randall Clark, Vacslav Glukhov, Georgy Subbotin, Maxim Nurgaliev, Aleksandr Kachkin, Lei Zeng, Dmitri M. Orlov

TL;DR

This work addresses the challenge of diagnosing plasma confinement state in fusion pilot plants under severe diagnostic constraints. It extends a previously ECE-based confinement-mode classifier by developing a Profile Reflectometer (PR)–based model and fusing the two via an ensemble to achieve higher accuracy and robustness, suitable for reactor environments. The PR classifier achieves 97% test accuracy, while the ensemble model reaches 99.2% test accuracy, with a robust analysis showing edge/pedestal information is key to H-mode detection. A chronological sliding-window robustness assessment indicates the ensemble remains the most reliable approach as operational data evolve, supporting practical deployment of PR-based diagnostics in future FPPs.

Abstract

As Fusion Pilot Plants (FPPs) are increasingly viewed as within reach, many engineering challenges remain. Not many diagnostics are expected to be available in a reactor environment. Survivability, maintainability, and limited port space substantially restrict the number of FPP-relevant diagnostics. One remaining challenge is developing tools and devices to extract plasma state information necessary for controlling an FPP from a limited subset of diagnostics. This work is part of an overarching project to address this challenge. The specific diagnostic subset to be used in FPPs is still under debate. We take the approach of developing machine-learning-based tools for different significant plasma state parameters, using already known FPP-viable diagnostics. Previously we developed a plasma confinement mode classifier utilizing the Electron Cyclotron Emission (ECE) diagnostic. Here, we expand on this by developing a Profile Reflectometer (PR) based classifier with 97\% test accuracy, and an ensemble model that combines the ECE and PR models into a single model, achieving 99\% test accuracy.

Plasma Confinement State Classification in Fusion Power Plants: Profile Reflectometer and Ensemble Diagnostics

TL;DR

This work addresses the challenge of diagnosing plasma confinement state in fusion pilot plants under severe diagnostic constraints. It extends a previously ECE-based confinement-mode classifier by developing a Profile Reflectometer (PR)–based model and fusing the two via an ensemble to achieve higher accuracy and robustness, suitable for reactor environments. The PR classifier achieves 97% test accuracy, while the ensemble model reaches 99.2% test accuracy, with a robust analysis showing edge/pedestal information is key to H-mode detection. A chronological sliding-window robustness assessment indicates the ensemble remains the most reliable approach as operational data evolve, supporting practical deployment of PR-based diagnostics in future FPPs.

Abstract

As Fusion Pilot Plants (FPPs) are increasingly viewed as within reach, many engineering challenges remain. Not many diagnostics are expected to be available in a reactor environment. Survivability, maintainability, and limited port space substantially restrict the number of FPP-relevant diagnostics. One remaining challenge is developing tools and devices to extract plasma state information necessary for controlling an FPP from a limited subset of diagnostics. This work is part of an overarching project to address this challenge. The specific diagnostic subset to be used in FPPs is still under debate. We take the approach of developing machine-learning-based tools for different significant plasma state parameters, using already known FPP-viable diagnostics. Previously we developed a plasma confinement mode classifier utilizing the Electron Cyclotron Emission (ECE) diagnostic. Here, we expand on this by developing a Profile Reflectometer (PR) based classifier with 97\% test accuracy, and an ensemble model that combines the ECE and PR models into a single model, achieving 99\% test accuracy.
Paper Structure (14 sections, 4 equations, 6 figures, 4 tables)

This paper contains 14 sections, 4 equations, 6 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Diagnostics
  3. The Profile Reflectometer Diagnostic
  4. The Electron Cyclotron Emission Diagnostic
  5. Data Analysis
  6. Hand Label Generation Process
  7. Peculiarities of Profile Reflectometer data
  8. The Profile Reflectometer Classifier
  9. The Model
  10. Model Analysis
  11. The Ensemble Model
  12. Robustness to Future Data
  13. Conclusion and future work
  14. Acknowledgment

Figures (6)

  • Figure 1: Distribution of the PR data used in training and testing according to the time within the shot, and the $\rho$ range.
  • Figure 2: A t-SNE heuristic visualization of PR labeled data. Apart from a few outliers at the boundary between L- and H-mode, the hand-labeled data points separate cleanly.
  • Figure 3: This plot is an example of how well the splines fit to the data, how they smooth out the noise, and how they perform the constant extrapolation in the H-mode case where the PR density limit is reached. The dashed vertical lines represent locations where density values will be taken for input to the GBC model.
  • Figure 4: This flow chart shows the full PR model, how data from the PR is processed and used for confinement mode classification.
  • Figure 5: This plot depicts the relative importance of each input into the GBC as determined by the location of the density measurement along $\rho$. This result shows the dominant contribution of input data from the edge region, strongly indicating the importance of pedestal detection for identifying H-mode. The physics that informed the construction of this model is identifiable in this plot.
  • ...and 1 more figures