Development Of The Plasma Disruption Prediction And Disruption Mitigation System On J-TEXT

Main disruption is a catastrophic event for tokamak operation which can’t be avoided completely. It can release stored energy of the plasma in several milliseconds. The so called thermal quench and current quench could cause great damage to the device as follows: large local thermal load, strong electromagnetic stress and runaway current with hundreds of mega-electron-volt(Me V) energy. Thus, prediction, avoidance and mitigation of disruption are high priority tasks. Development of the plasma disruption prediction and mitigation system on J-TEXT could help accumulating relevant technical experience, and can also provide some meaningful base for other tokamaks.This thesis firstly establishes the disruption database on J-TEXT and analyzes the main causes and possible mechanisms leading to disruption, then we divide these disruptions into several types for further investigations. For density limit disruptions, we adopt artificial neural network as the main modeling tool aiming to use multi-parameters, and conduct a systematic research. An optimization model containing multiple relevant diagnostic signals of density limit is established. It is found that this model can predict about 80 percent of density limit disruptions, and the prediction can be made up to 5ms prior to the disruption time. Furthermore, this model is not only suitable for disruption prediction but also for density limit disruption avoidance. Through sensitivity analysis of the model, we can make reasonable interpretations about the phenomenon and mechanisms of density limit disruptions on J-TEXT. We also simply analyze the influences of resonant magnetic perturbation(RMP) on the tearing mode instabilities and disruption. At last, this thesis proposed an optimized cascade neural network model. The results presented show that by properly training the newly proposed model can reduce the false alarm rate apparently under the premise of ensuring the success rate at about 80 percent.The fundamental purpose of disruption prediction is to implement avoidance or mitigation strategy, massive gas injection(MGI) and shattered pellets injection are the main mitigation methods. In order to meet the requirements of disruption physical experiments, two MGI hardware systems have been established on J-TEXT. We have analyzed and introduced the structure and parameters design of the MGI valves, principles of the pulse power supply and gas path of the MGI hardware systems. We have also designed an experimental disruption mitigation control system for the two MGI valves. The experimental control system has been applied and optimized after several rounds of experiments and performs well, a series of meaningful experimental results have been obtained.