Description
Efficient plasma heating is a crucial prerequisite for achieving fusion reactions in KSTAR, where the plasma temperature must reach the order of 100 million degrees Celsius. Among various heating mechanisms, electron cyclotron heating (ECH) utilizes the electron cyclotron resonance frequency of RF waves to transfer energy directly to the electrons. However, in practical operation, the actual RF power absorbed by electrons may fall short of theoretical predictions, as evidenced by zero-dimensional (0-D) power efficiency analyses. Additionally, the spatial distribution of absorbed power varies radially within the plasma, necessitating a more detailed one-dimensional (1-D) power profile analysis. Ray tracing simulations [1] serve as an essential tool to model and evaluate these heating profiles, capturing the complex interactions between RF waves and the plasma environment. However, the efficiency of plasma heating and the resultant radial heating profile are governed by multiple environmental parameters, exhibiting highly nonlinear relationships. In this study, we leverage machine learning techniques to establish predictive models for these relationships, aiming to enhance the estimation of radial heating profiles and optimize the performance of ECH systems. Our approach enables accelerated analysis of heating efficiency under various plasma conditions, contributing to the refinement of KSTAR’s operational strategies for improved fusion performance.
References
[1].A.P. Smirnov, R.W. Harvey, and K. Kupfer, A general ray tracing code GENRAY, Bull Amer. Phys. Soc. Vol 39, No. 7, p. 1626 Abstract 4R11 (1994)
