-
On learning latent dynamics of the AUG plasma state
Authors:
A. Kit,
A. E. Järvinen,
Y. R. J. Poels,
S. Wiesen,
V. Menkovski,
R. Fischer,
M. Dunne,
ASDEX-Upgrade Team
Abstract:
In this work, we demonstrate the utility of state representation learning applied to modeling the time evolution of electron density and temperature profiles at ASDEX-Upgrade (AUG). The proposed model is a deep neural network which learns to map the high dimensional profile observations to a lower dimensional state. The mapped states, alongside the original profile's corresponding machine paramete…
▽ More
In this work, we demonstrate the utility of state representation learning applied to modeling the time evolution of electron density and temperature profiles at ASDEX-Upgrade (AUG). The proposed model is a deep neural network which learns to map the high dimensional profile observations to a lower dimensional state. The mapped states, alongside the original profile's corresponding machine parameters are used to learn a forward model to propagate the state in time. We show that this approach is able to predict AUG discharges using only a selected set of machine parameters. The state is then further conditioned to encode information about the confinement regime, which yields a simple baseline linear classifier, while still retaining the information needed to predict the evolution of profiles. We then discuss the potential use cases and limitations of state representation learning algorithms applied to fusion devices.
△ Less
Submitted 15 December, 2023; v1 submitted 28 August, 2023;
originally announced August 2023.
-
Parametric scaling of power exhaust in EU-DEMO alternative divertor simulations
Authors:
A. E. Järvinen,
L. Aho-Mantila,
T. Lunt,
F. Subba,
G. Rubino,
L. Xiang
Abstract:
Investigations of parametric scaling of power exhaust in the alternative divertor configuration (ADC) SOLPS-ITER simulation database of the EU-DEMO are conducted and compared to predictions based on the Lengyel model. The Lengyel model overpredicts the necessary argon concentrations for LFS divertor detachment by about a factor of 5-10 relative to the SOLPS-ITER simulations. Therefore, while the L…
▽ More
Investigations of parametric scaling of power exhaust in the alternative divertor configuration (ADC) SOLPS-ITER simulation database of the EU-DEMO are conducted and compared to predictions based on the Lengyel model. The Lengyel model overpredicts the necessary argon concentrations for LFS divertor detachment by about a factor of 5-10 relative to the SOLPS-ITER simulations. Therefore, while the Lengyel model predicts that plasmas with accetable divertor heat loads in EU-DEMO would exceed the tolerable upstream impurity concentrations by a large margin, there are several SOLPS-ITER solutions within an acceptable operational space. The SOLPS-ITER simulations indicate that, unlike assumed by the standard Lengyel model, there are significant heat dissipation mechanisms other than argon radiation, such as cross-field transport, that reduce the role of argon radiation by a factor of 2 to 3. Furthermore, the Lengyel model assumes that the radiation front is powered by parallel heat conduction only, which tends to lead to a narrow radiation front as the radiative efficiency increases strongly with reducing thermal conductivity. As a result, the radiative volume and total impurity radiation are suppressed for a given impurity concentration. However, the SOLPS-ITER simulations indicate that other mechanisms, such as cross-field transport, can compete with parallel heat conduction within the radiative front and increase the radiative volume. Due to these findings, usage of the standard Lengyel model for analyzing scaling between divertor conditions and configurations for devices such as EU-DEMO is strongly discouraged.
△ Less
Submitted 16 September, 2022;
originally announced September 2022.
-
Measuring the Electron Temperature and Identifying Plasma Detachment using Machine Learning and Spectroscopy
Authors:
C. M. Samuell,
A. G. Mclean,
C. A. Johnson,
F. Glass,
A. E. Jaervinen
Abstract:
A machine learning approach has been implemented to measure the electron temperature directly from the emission spectra of a tokamak plasma. This approach utilized a neural network (NN) trained on a dataset of 1865 time slices from operation of the DIII-D tokamak using extreme ultraviolet / vacuum ultraviolet (EUV/VUV) emission spectroscopy matched with high-accuracy divertor Thomson scattering me…
▽ More
A machine learning approach has been implemented to measure the electron temperature directly from the emission spectra of a tokamak plasma. This approach utilized a neural network (NN) trained on a dataset of 1865 time slices from operation of the DIII-D tokamak using extreme ultraviolet / vacuum ultraviolet (EUV/VUV) emission spectroscopy matched with high-accuracy divertor Thomson scattering measurements of the electron temperature, $T_e$. This NN is shown to be particularly good at predicting $T_e$ at low temperatures ($T_e < 10$ eV) where the NN demonstrated a mean average error of less than 1 eV. Trained to detect plasma detachment in the tokamak divertor, a NN classifier was able to correctly identify detached states ($T_e<5$ eV) with a 99% accuracy (F$_1$ score of 0.96) at an acquisition rate $10\times$ faster than the Thomson scattering measurement. The performance of the model is understood by examining a set of 4800 theoretical spectra generated using collisional radiative modeling that was also used to predict the performance of a low-cost spectrometer viewing nitrogen emission in the visible wavelengths. These results provide a proof-of-principle that low-cost spectrometers leveraged with machine learning can be used both to boost the performance of more expensive diagnostics on fusion devices, and be used independently as a fast and accurate $T_e$ measurement and detachment classifier.
△ Less
Submitted 21 October, 2020;
originally announced October 2020.
-
Role of Poloidal $\mathbf{E}\times\mathbf{B}$ Drift in Divertor Heat Transport in DIII-D
Authors:
A. E. Järvinen,
S. L. Allen,
A. W. Leonard,
A. G. McLean,
A. L. Moser,
T. D. Rognlien,
C. M. Samuell
Abstract:
Simulations for DIII-D high confinement mode plasmas with the multifluid code UEDGE show a strong role of poloidal $\mathbf{E}\times\mathbf{B}$ drifts on divertor heat transport, challenging the paradigm of conduction limited scrape-off layer (SOL) transport. While simulations with reduced drift magnitude are well aligned with the assumption that electron heat conduction dominates the SOL heat tra…
▽ More
Simulations for DIII-D high confinement mode plasmas with the multifluid code UEDGE show a strong role of poloidal $\mathbf{E}\times\mathbf{B}$ drifts on divertor heat transport, challenging the paradigm of conduction limited scrape-off layer (SOL) transport. While simulations with reduced drift magnitude are well aligned with the assumption that electron heat conduction dominates the SOL heat transport, simulations with drifts predict that the poloidal convective $\mathbf{E}\times\mathbf{B}$ heat transport dominates over electron heat conduction in both attached and detached conditions. Since poloidal $\mathbf{E}\times\mathbf{B}$ flow propagates across magnetic field lines, poloidal transport with shallow magnetic pitch angles can reach values that are of the same order as would be provided by sonic flows parallel to the field lines. These flows can lead to strongly convection dominated divertor heat transport, increasing the poloidal volume of radiative power front, consistent with previous measurements at DIII-D. Due to these convective flows, the Lengyel integral approach, assuming zero convective fraction, is expected to provide a pessimistic estimate for radiative capability of impurities in the divertor. For the DIII-D simulations shown here, the Lengyel integral approach underestimates the radiated power by a factor of 6, indicating that for reliable DIII-D divertor power exhaust predictions, full 2D calculations, including drifts, would be necessary.
△ Less
Submitted 19 September, 2019;
originally announced September 2019.
