Data-driven model discovery for plasma turbulence modelling

An important problem in nuclear fusion plasmas is the prediction and control of turbulence which drives cross-field transport, thus leading to energy loss from system deteriorating confinement. Turbulence, being a highly nonlinear multiscale process, challenging theoretically describe computationally model. Most advanced computational models fall into one two categories: fluid or gyro-kinetic. They both come at high cost cannot be applied for routine simulation plasma discharge evolution control. Development reduced based on (physics informed) artificial neural networks could potentially fulfil need affordable simulations turbulence. However, training requires an extensive data base obtained lack extrapolation capability scenarios not originally encountered during training. This leads limited validity may prove adequate predicting future machines. In contrast, we explore data-driven model discovery approach sparse regression infer governing partial differential equations directly data. Our input are generated by drift-wave according Hasegawa–Wakatani modified models. Balancing accuracy complexity enables reconstruction systems accurately describing dynamics simulated sets. Sparse hungry can extrapolated unexplored parameter ranges. We demonstrate potential this modelling. The findings show that methodology promising development efficient as well existing cross-validation.