Recognition of Real-Time Dynamic Parameters of Vertical Displacement Based on Neural Network

Abstract

Vertical position control is crucial for stabilizing the plasma with elongated configurations. Currently, a number of model-based control algorithms, in which the control parameters are designed based on the vertical position response parameters, have been established in different devices and become the solution of future fusion devices’ vertical position stabilization. The vertical response parameters can be calculated from the given scenario with the physical-based response model before the experiments. However, the plasma vertical position’s response varies with changes in plasma configuration, resulting in different control parameter requirements. A real-time estimation of the vertical position response can significantly improve the capability and robustness of the vertical position control. In this article, the neural network (NN) is used to identify the dynamic vertical displacement response parameters in real time. Experiments demonstrate that the NNs accuracy and real-time performance meet the control requirements and would enhance control capabilities.