Research On Nolinear Model Predictive Iterative Learning Control

Iterative learning control (ILC) and Model predictive control (MPC) are both popular algorithms which are gradually developed and widely used in industrial processes control. The generation and research status of ILC and MPC are introduced in the first place. Then, it analyzes the weaknesses s of the above sole algorithm. Thus, the integration of MPC and ILC is proposed to constitute the model predictive iterative learning control (MPILC). Most industrial processes are highly nonlinear systems, By learning the weaknesses of the existing MPILC, three types of modified nonlinear model predictive iterative learning control(NMPILC) based on T-S fuzzy model are proposed. Based on each local model, a local MPILC controller is designed. The global controller is characterized by a weighted sum of the focal controllers according to the actual operating condition changes. The local MPILC of type I NMPILC modified the three aspects from the classical MPILC, The convergence property has been demonstrated. A PWR nuclear power plant is adopted to illustrate the performance of the NMPILC. The type II NMPILC proposes a feed forward-feedback control law in which ILC acts as a feedforward controller and provides most of control signals, MPC acts as a feedback controller and compensates for uncertainties and real-time disturbances, a simulation of PH neutralization process is adopted to illustrate the performance. Considering a nonlinear system with repetitive disturbances, the type III NMPILC combines the predictive function control (PFC) and ILC to constitute the Nonlinear Iterative Learning Predictive Function Control (NPFILC), online computing time is deduced and simulation results show the performance the the NPFILC.