Pedestal Pressure Prediction via Mode-Specific Regression in Tokamak Plasmas

Obtaining an accurate prediction of the pedestal pressure in tokamak plasmas is integral for expanding understanding of confinement regimes and optimizing performance in magnetically confined plasmas in fusion devices. This work employs a data-driven approach for pedestal pressure prediction using power-law regression models trained on DIII-D data. The pressure is always predicted at a normalized toroidal radius of ρ = 0.9. Input parameters include plasma current, toroidal magnetic field, elongation, triangularity, electron density, and normalized beta. Parameterization of the plasma shape is the same as used in the EPED model. One of the novel features of this work is that a single power law linear regression can describe various regimes. Separate regression models were developed for L-mode (R2 = 0.88), H-mode (R2 = 0.88), and combined regimes (R2 = 0.92) where the pedestal height can continuously vary from L- to H- mode values without a priori filtering. Most H-mode data included edge-localized mode (ELM) plasma instabilities. Future work could investigate models for ELM-suppressed regimes. Additional and separate models were developed for plasmas in negative triangularity (R2 = 0.84) and a combination of L-mode, H-mode, and negative triangularity (R2 = 0.92). The final implementation allows users to predict the pedestal pressure given the input parameters. This tool provides a fast estimate of pedestal conditions.

Work supported by US DOE under DE-FC02-04ER54698.