Nonlinear Predictive Functional Control Based on Hopfield Network and its Application in CSTR
Peng Guo · 2006
CSTR is a nonlinear chemical reactor widely used in chemical industry and can be simplified as an affine nonlinear system. Hopfield network is a neural network with rich dynamic characteristics. In this paper, affine nonlinear system is treated as black box, and is identified with Hopfield network. After obtaining the relative degree of the nonlinear system from the network, state feedback linearization method is used to transform CSTR to a one-order linear system. The state variables and Lie derivatives needed in the transform can be obtained from the Hopfield network. Finally, a PFC controller is designed to control the linear system. Simulations prove that the new method has good control performance