On parallel structure RC for better performance and NN based tilc for a class of nonlinear systems

Repetitive control and iterative learning control are both learning control schemes which update and refine control sequence using tracking errors in past trials.With the superior ability to handle system uncertainty and repetitive disturbance, repetitive control and iterative learning control have been applied to industry successfully in the past decades.The basic ideas of these two control schemes are similar to each other, but their analysis methods and applications are quite different.This thesis is divided into two parts , each of which focuses on the developments of repetitive control and iterative learning control, respectively.In PART I, a new parallel structure repetitive control scheme is proposed.This new repetitive controller realizes selective compensation on the targeted harmonics.Compared with the conventional methods, it achieves better tracking performance with less delay and computational burden.Furthermore, correction factors which modify the poles of conventional repetitive control in fractional cases are introduced.As a result, robust response is achieved in fractional cases where sampling frequency is not integer multiple of fundamental frequency.As an application, a high-performance grid simulator controlled by parallel structure fractional repetitive controller is designed and implemented.Experimental results testify the effectiveness of the proposed control scheme.In PART II, combination of neural network and iterative learning control is exploited to address system uncertainties and zero-error initial condition in nonlinear