An accurate evolution model is crucial for effective control and in-depth study of fusion plasmas. Physics-based evolution models often encounter challenges such as insufficient robustness or excessive computational costs. Given the proven strong fitting capabilities of deep learning methods across various domains, including plasma research, this paper introduces a deep learning based magnetic measurement evolution method named PaMMA-Net (Plasma Magnetic Measurements Incremental Accumulative Prediction Network). This model is capable of evolving magnetic measurements in tokamak discharge experiments within 1000 ms with a step of one millisecond. In contrast to directly evolving specific equilibrium parameters, magnetic measurements evolution is trained on precise experimental measurements, thereby circumventing errors in data processing. Furthermore, equilibrium reconstruction based on the evolution of magnetic measurements could yield a more comprehensive set of equilibrium parameters, including plasma shape, current center, etc. Leveraging an incremental prediction approach and data augmentation techniques tailored for magnetic measurements, PaMMA-Net achieves superior evolution results compared to existing studies. The tests conducted on real experimental data from EAST validate the high generalization capability of the proposed method.
This paper introduces PaMMA-Net, a deep learning model that can accurately predict the evolution of magnetic measurements in fusion plasma experiments. Unlike traditional physics-based models, PaMMA-Net learns directly from experimental data, making it more robust and efficient. It can predict magnetic measurements up to 1000 ms in advance, enabling better plasma control and analysis.