The work builds on research to develop an integrated control and actuator sharing framework for standard controllers at tokamak à configuration variable (variable configuration tokamak; TCV) at the École Polytechnique Fédérale de Lausanne (EPFL).
End-to-end implementation: PACMAN is organized into four blocks, taking the process from diagnostic data collection through post-processing of automated adjustments made on the basis of the model calculations.
The process begins with real-time measurements of what’s happening in the tokamak, including temperatures, densities, and magnetic signals. It feeds this data to a range of ML models.
According to the paper, PACMAN was designed to be modular and flexible. This allows it to support a large variety of models, the intention being that many existing and potential future models and controllers with different functions can be unified under its framework. The researchers believe these characteristics will also make it easy to adapt PACMAN for use to other tokamaks beyond DIII-D.
“PACMAN uses a flexible setup where building-block AI algorithms can be put together. You can add a new one, swap one out or run several at once without touching the rest of the system,” said Egemen Kolemen, associate professor of mechanical and aerospace engineering at Princeton University, jointly appointed with the Andlinger Center for Energy and the Environment and PPPL. “That modularity is what turns AI plasma control from a series of one-off demonstrations into infrastructure the whole fusion community can build on.”
Currently, PACMAN includes ML surrogates of physics codes, event prediction models, plasma profile prediction models, and more.
The model outputs get sent to the controller block, which makes decisions about how to translate the information into actionable steps for controlling the plasma, with user-set minimums and maximums to avoid unwanted or unsafe controller behavior. These get sent to actuators, which carry out the adjustments.
“The whole PACMAN framework typically runs in about 20 milliseconds, and it’s not running once. It’s running again and again and again. It can see small things happening in the plasma and adjust in a way that a human would never be able to do,” said Andy Rothstein, a graduate student in Princeton University’s Department of Mechanical and Aerospace Engineering. “We developed this framework so that models could communicate, outputs from those models could be shared and we could do exciting physics in one integrated system.”
Experimental tests: The PPPL team implemented PACMAN at DIII-D and tested a handful of scenarios.
According to PPPL, PACMAN was able to take complete control of DIII-D’s six gyrotrons, predict tearing modes and sudden bursts of energy from the plasma’s edge, detect and control waves in the plasma driven by fast particles, and adjust the plasma’s density and rotation.
The ability to predict the plasma’s behavior very quickly can lead to significant performance improvements; for example, conventional controllers can only detect a tearing mode after it has started, so by the time actions to suppress it kicks in, the system is already playing catchup in a way that can lead to a lot of performance degradation.
“Machine learning models can describe the plasma behavior very well, and importantly, they are the only way we have to model the plasma in millisecond times. The speed of these models is what’s key for control,” said Hiro Farre Kaga, a graduate student in the Princeton Program in Plasma Physics. “In one of the experiments we present, a machine learning model predicts the tearing mode about 200 milliseconds in advance, so the plasma can be changed to avoid it in the first place.”
.jpg)
