Polynomial approach to nonlinear predictive GMV control

A simple approach is described for the design of nonlinear predictive controllers. The Nonlinear Predictive Generalized Minimum Variance (NPGMV) control algorithm is introduced for nonlinear discrete-time multivariable systems. The system is represented by a combination of a stable nonlinear subsystem where no structure is assumed and a linear subsystem that may be unstable and modelled in polynomial matrix form. The multi-step predictive control cost index to be minimised involves both weighted error and control signal costing terms. The solution for the control law is derived in the time-domain using a very nonlinear operator model of the process. The controller includes an internal nonlinear model of the process but because of the assumed structure of the system, that has a linear disturbance model, the polynomial equations for the predictor are linear.