Implementation of EAST Fast Vertical Position Control With Neural Network-Based Estimator

The vertical instability of the plasma is one of the main causes of large disruption. For the operation of an elongated tokamak plasma, it is necessary to control the vertical instability of the plasma effectively. This article implements a new vertical position control algorithm by using a neural network model for more accurate vertical position estimation. Meanwhile, a high-speed acquisition and fast control system is designed, which can reduce latency and provide sufficient computing resources for the neural network. Besides, the control algorithm is deployed in the new system with a separate server. Finally, the benefit of using a neural network in the new control system to control vertical position is confirmed by EAST experiments. Experimental results indicate that the neural network can accurately calculate the vertical position under different configurations, and the high-speed acquisition and fast control system reduces the response transmission delay and enhances the vertical position control.