Predictive Fuzzy Control Using Particle Swarm Optimization for Magnetic Levitation System

Abstract

AbstractThe magnetic levitation system (MLS) has the characteristics of time-delay, open-loop unstable and non-linear. Considering the poor robustness and ability to tack ideal position of conventional algorithms, a predictive fuzzy proportional-integral-derivative (PID) with particle swarm optimization (PSO-PFPID) controller was proposed. With the recursive least squares (RLS) algorithm, a controlled auto-regressive integrated moving-average (CARIMA) model is identified on line and it serves for the predictive model of the generalized predictive control (GPC). Then the predicted optimal control law in the future time is served as the input of fuzzy PID (FPID). To enhance the dynamic and steady performance of the predictive fuzzy PID (PFPID) controller at the same time, the softening coefficient α and forgetting factor μ of the PFPID controller are globally optimized offline with the particle swarm optimization (PSO) algorithm. Finally, the maglev ball system is employed as the controlled object of the simulation and experiment and the mathematical model in balance points at the equilibrium point. Compared with PID, cascade GPC PID(PPID), simple PFPID, the PSO-PFPID controller has better robustness and make the controlled object stable in the case of mismatch. It can effectively adapt to system parameters changes.KeywordsFPID controllerGeneralized predictive PID controlMaglev systemParameter identificationPSO algorithmRobustness