Neural-Network Model of the Multi-Mode Anomalous Transport Module
Real-time-capable Machine-learning-based Accelerated Model of MMM for Transport Simulations
S. Morosohk, T. Rafiq, E. Schuster
28th IAEA Fusion Energy Conference
Nice, France, October 12-17, 2020 -> May 10-15, 2021 (Remote)
Abstract
|
|
A neural network version of the Multi-Mode Anomalous Transport Module,
known as MMMnet, has been developed to calculate plasma turbulent
diffusivities in DIII-D with a calculation time suitable for control
applications. The current version of MMMnet uses a simple artificial
neural network structure to predict the electron thermal, ion thermal,
impurity particle, and toroidal momentum diffusivities while reproducing
MMM data with 86% accuracy and keeping the calculation time as a fraction
of that associated with MMM. A new version of MMMnet is being trained
to add ion particle and poloidal momentum diffusivities as additional
outputs. Model-based control techniques require models with fast
calculation times, making many existing physics-oriented predictive
codes unsuitable. The control-oriented predictive code COTSIM
(Control-Oriented Transport Simulator) has been developed to calculate
the most significant plasma dynamics in response to the different
actuators while running at a speed useful for control design. In order
to achieve this calculation speed, COTSIM relies on scaling laws and
control-level models, often leading to a lower range of validity across
different scenarios and less accurate predictions. Replacing some of
these scaling laws and control-level models with neural network versions
of more complex physics-level models has the potential of increasing
the range of validity and the level of accuracy of COTSIM without
compromising its computational speed. In this work, MMMnet is integrated
into COTSIM to improve the turbulent diffusivity predictions, which
will in turn improve the prediction accuracy associated with the dynamics
of many plasma properties.