Estimation of ECH power deposition based on neural networks and fuzzy logic in plasma fusion Tokamaks

In order to stabilize magnetic hydro dynamics (MHD) activity in a Tokamaks, the measurement data acquired by different sensors along with prior information obtained from predictive plasma models are used. Suppression of plasma instabilities is a key issue to improve the confinement time of controlled thermonuclear fusion with Tokamaks. This paper proposes a method based on Self Organizing Maps (SOM) type Neural Network to estimate the Electron Cyclotron Heating (ECH) power deposition radius (r_DEP) during plasma confinement. The proposed approach that is a part of the control system to stabilize MHD instability, has been compared to the Bayesian filter approach which has been proposed previously. The Bayesian approach uses on-line information acquired from Electron Cyclotron Emission (ECE) sensors and prior information got from ray-tracing code to compute the mean and standard deviation of the estimated deposition channel. The SOM approach mostly relies on ECE sensors data instead of prior information and tries to estimate the power deposition channel in real-time with less computations. A fuzzy system is also designed to reduce the uncertainty of the SOM algorithm. These algorithms have been fully compared in different aspects too. The algorithms have been tested on off-line ECE channels data, obtained from an experimental shot at Frascati Tokamak Upgrade (FTU), Frascati, Italy.