Predictive Control

8Predictive Control for Nonlinear Systems

Predictive control algorithms were firstly proposed for linear systems. The predictive control algorithms that appeared early on, such as DMC, MAC, GPC, etc. are all based on linear models, where the future outputs are predicted according to the proportion and superposition properties of linear systems. If the plant is weakly nonlinear, it can be approximated by a linear model and the linear predictive control algorithm can be used. In this case, the degeneration of the system performance caused by model mismatch due to nonlinearity is not critical and can be compensated to some extent by introducing appropriate feedback correction. However, if the plant is strongly nonlinear, the output prediction by using a linear model may lead to a large deviation from the actual one and thus the control result may be far from that predicted with the linear model. This means that it cannot be simply handled by using the linear predictive control algorithm. For predictive control applications, how to develop efficient predictive control strategies and algorithms with respect to the characteristics of nonlinear systems has always been the focus of attention. In this chapter, we firstly give the general description of predictive control for nonlinear systems and then introduce some representative strategies and methods. The predictive control algorithms for nonlinear systems are not unified and exhibit a trend of diversification according to different ...