Improving feedforward disturbance compensation capabilities in Generalized Predictive Control

Abstract

This paper deals with the measurable disturbance rejection problem in Generalized Predictive Control (GPC). First, it is analyzed how the unconstrained GPC algorithm with implicit disturbance compensation can be interpreted as a typical feedback plus feedforward control scheme, where the main feature is that the feedforward action includes future estimations of the measurable disturbances. Then, it is shown that classical GPC cannot always eliminate the effect of measurable disturbances even using perfect disturbance models and having exact disturbance estimations along the prediction horizon. To overcome this problem a particular GPC tuning condition is proposed, which allows the improved GPC controller to eliminate the disturbance effect even in those cases where causality and instability problems can appear in the relation between the dynamics of the load disturbance and the process output with the dynamics of the control signal and the process output. Since the new tuning condition for disturbance compensation in GPC leads to a high bandwidth in the feedback loop, a two degrees of freedom control scheme within the Filtered Smith Predictor (FSP)-based GPC framework has been implemented to improve the robustness capabilities of the control law. Simulation examples are presented to show the main advantages of the proposed control scheme, including a realistic simulation based on a greenhouse climate control problem where estimators for the main process disturbances are also designed.