Experimental implementation of MIMO model predictive controller-based second order divided difference filter for nonlinear systems

摘要

In this paper, we propose an experimental implementation of a nonlinear predictive controller based on the second order divided difference filter (DDF2) which is a derivate-free estimator. The used state estimator avoids the determination of Jacobian matrices required with the extended Kalman filter (EKF) for an easy implementation with nonlinear systems. In addition, the second order Stirling's interpolation of the estimator allows a more accurate a priori state prediction. A constrained nonlinear predictive controller-based DDF2 is designed to control a multivariable three-tank system. In order to generate the control signals, the controller solves a nonlinear objective function. To overcome the offset output error that appeared due to the inaccuracy of the system model, an internal state model correction scheme is adopted at each iteration. Practical results show the reliability of the proposed method in state estimation, setpoint tracking, and smooth control signal generation. The proposed controller outperforms the model predictive control based on linear state space model derived from the linearisation of the system model around the same setpoint trajectory.