Nuclear Fusion and Artificial Intelligence: the Dream of Limitless Energy

Ever since the 1930s when scientists, namely Hans Bethe, discovered that nuclear fusion was possible, researchers strived to initiate and control fusion reactions to produce useful energy on Earth. The best example of a fusion reaction is in the middle of stars like the Sun where hydrogen atoms are fused together to make helium releasing a lot of energy that powers the heat and light of the star. On Earth, scientists need to heat and control plasma, an ionised state of matter similar to gas, to cause particles to fuse and release their energy. Unfortunately, it is very difficult to start fusion reactions on Earth, as they require conditions similar to the Sun, very high temperature and pressure, and scientists have been trying to find a solution for decades. 

In May 2019, a workshop detailing how fusion could be advanced using machine learning was held that was jointly supported by the Department of Energy Offices of Fusion Energy Science (FES) and Advanced Scientific Computing Research (ASCR). In their report, they discuss seven 'priority research opportunities':

  • 'Science Discovery with Machine Learning' involves bridging gaps in theoretical understanding via identification of missing effects using large datasets; the acceleration of hypothesis generation and testing and the optimisation of experimental planning. Essentially, machine learning is used to support and accelerate the scientific process itself.

  • 'Machine Learning Boosted Diagnostics' is where machine learning methods are used to maximise the information extracted from measurements, systematically fuse multiple data sources and infer quantities that are not directly measured. Classifcation techniques, such as supervised learning, could be used on data that is extracted from the diagnostic measurements. 

  • 'Model Extraction and Reduction' includes the construction of models of fusion systems and the acceleration of computational algorithms. Effective model reduction can result in shorten computation times and mean that simulations (for the tokamak fusion reactor for example) happen faster than real-time execution.

  • 'Control Augmentation with Machine Learning'. Three broad areas of plasma control research would benefit significantly from machine learning: control-level models, real-time data analysis algorithms; optimisation of plasma discharge trajectories for control scenarios. Using AI to improve control mathematics could manage the uncertainty in calculations and ensure better operational performance.

  • 'Extreme Data Algorithms' involves finding methods to manage the amount and speed of data that will be generated during the fusion models.

  • 'Data-Enhanced Prediction' will help monitor the health of the plant system and predict any faults, such as disruptions which are essential to be mitigated.

  • 'Fusion Data Machine Learning Platform' is a system that can manage, format, curate and enable the access to experimental and simulation data from fusion models for optimal usability when used by machine learning algorithms.

Data science methods from the fields of machine learning and artificial intelligence (ML/AI) offer opportunities for enabling or accelerating progress toward the realization of fusion energy by maximizing the amount and usefulness of information extracted from experimental and simulation output data.
— U.S. Department of Energy
The European Joint European Torus (JET) Tokamak Fusion Reactor during (right) and after operation. (Source: Culham Centre for Fusion Energy, JET)

The European Joint European Torus (JET) Tokamak Fusion Reactor during (right) and after operation. (Source: Culham Centre for Fusion Energy, JET)

The Princeton Plasma Physics Lab (PPPL) has also used AI in their Fusion Recurrent Neural Network which aims to predict plasma disruptions ("a fast and anomalous loss of stability that can cause severe damage to plasma facing components"). The prediction system is led by William Tang, a professor in the Department of Astrophysical Sciences at Princeton and a principal research physicist at PPPL) and is considered to be "the first machine learning disruption predictor capable of consistently outperforming, on all metrics that matter, a simple “locked-mode” based predictor." In Eurofusion's paper on their advanced disruption predictor, they explain what the locked-mode is and how it is used in prediction: "When macroscopic instabilities start locking to the wall, the amplitude of the signal used to detect them (called locked mode) grows during the slowing down of their rotation. Therefore, the locked mode amplitude is routinely used as precursor of disruptions caused by this locking of instabilities to the wall.".

One company working towards making nuclear fusion commercial is TAE Technologies (formerly known as Tri-Alpha Energy) and its CEO Michl Binderbauer claims that "commercialisation is coming in the next five years". Google Research's Applied Science branch has helped TAE discover new fusion techniques and CEO John Platt has said that "fusion has this potential for unlimited energy". Even though Binderbauer's claim was met with scepticism from the scientific community, AI could be the game-changer when it comes to nuclear fusion on Earth.

(Thumbnail Source: NASA/SDO/Seán Doran)