Neural network predictive control based on predicted steady state error

A radial basis function network (RBF network) is used as steady state model. A direct method of one\|dimensional optimization——golden section algorithm is used to seek the output of the controller. To improve the dynamic response of the system, we also add another RBF network used as dynamic model. Through the simulation of a nonlinear process——pH controlling process, we can see the good stability and dynamic response of the algorithm.