Edge Plasma Density Profile Estimations Using Bayesian Model and Gaussian process in the KSTAR Hydrogen Beam Emission Spectroscopy*

We have developed a Bayesian model for inferring edge plasma density profiles using the KSTAR hydrogen beam emission spectroscopy (H-BES) system. The system measures Doppler-shifted D-alpha emission and consists of 16 radial and 4 poloidal channels, each with approximately 1cm spatial resolution. The poloidal channels could provide an increased precision in radial profile measurement through the integration of equilibrium information. The intensities of the Doppler-shifted D-alpha lines are represented as a function of plasma density, guided by a multi-state model that encompasses the interactions of neutral deuterium beam atoms with plasma particles. The Gaussian process prior is employed to model density profiles and the posterior distribution is explored using a Markov Chain Monte Carlo (MCMC) method. Through the use of a Gaussian process and relative intensity calibration, the method can estimate instrument effects and absolute calibration factor. To provide a more accurate calculation of the absolute calibration factor, we have integrated this model with an edge interferometer. Given the growing significance of the 3D field in plasma control, the ability to simultaneously measure both the density profile and density fluctuations at a single toroidal position using a single diagnostic will significantly enhance our understanding of edge plasma physics.