AI Surrogate Modeling for Real-Time Tokamak Equilibrium Prediction: Benchmarking Neural Architectures and Validation on EXL-50U
Researchers have developed an AI surrogate framework for predicting plasma equilibrium in tokamaks. They tested five neural architectures on a large dataset and found that the CNN model achieved a good balance of accuracy, robustness, and speed. The surrogates were validated on a real device, the EXL-50U tokamak, with errors comparable to those from traditional methods. The study provides guidance for selecting reliable AI-based predictors for fusion applications.
Researchers have developed an AI surrogate framework for predicting plasma equilibrium in tokamaks. They tested five neural architectures on a large dataset and found that the CNN model achieved a good balance of accuracy, robustness, and speed. The surrogates were validated on a real device, the EXL-50U tokamak, with errors comparable to those from traditional methods. The study provides guidance for selecting reliable AI-based predictors for fusion applications.
---
Why it matters: This work matters because it addresses a critical challenge in fusion research: predicting plasma equilibrium in real-time. The development of accurate and efficient AI surrogates could enable more precise control and optimization of tokamak operations, which is essential for achieving stable and efficient fusion reactions.
Source: https://arxiv.org/abs/2608.23217
This article was originally published at: https://arxiv.org/abs/2608.23217