Avoidance of vertical displacement events in DIII-D using a neural network growth rate estimator

摘要: Abstract Robust disruption avoidance techniques are critical for the development of reliable fusion reactor devices. A viable will require non-disruptive, long pulse operation where simply shutting down a discharge is undesirable. To achieve such performance, plasma must be controlled to continuously avoid hazardous regimes instead asynchronously aborting. recent experiment on DIII-D demonstrated first time real-time control proximity disruptive instability boundary. In particular, vertical growth rate, an eigenvalue that characterizes degree plasma's position, was regulated so as not exceed DIII-D's controllability limit. The open-loop rate estimated in real system using neural network model trained with tens thousands shots. replicate results RZRIG [1] , rigid displacement code calculating rate. Once trained, producing estimate multiple orders magnitude faster than RZRIG, thereby making calculation suitable execution. by adjusting elongation and distance inner wall vessel, this regulation shown reliably event disruptions (i.e. uncontrolled oscillations) plasma. This work presents these experimental results, including dynamic performance effectiveness technique. Details presented training model, concerns hyperparameter tuning uncertainty quantification. Additionally, methodology embedding into discussed.