Machine learning for electrostatic plasma turbulence classification in tokamaks

Abstract Artificial intelligence driven by large data is becoming increasingly important in magnetic fusion research. Here, we scan the plasma gradient space for cyclone base case parameters using nonlinear gyrokinetic simulations to generate data for typical electrostatic drift wave turbulences. The main candidates, ion temperature gradient mode (ITG) and trapped electron mode (TEM) are then classified and labeled by conventional methods for the data sets in the linear stage. We then apply a classical machine learning algorithm, namely the support vector machine (SVM), and use plasma gradients or turbulent transport coefficients as the input to classify the type of drift wave turbulence. Simple distance formulae are derived for rapid classification of the turbulence type, demonstrating effectiveness suitable for future theoretical analyses and real-time experimental applications.