A filtered tuning method for a GPC controller

This paper presents a new tuning method based on model parameters identified in closed-loop. For classical controllers such as PI(D) controllers a large number of simple tuning methods for various application areas exist. However, when it comes to designing a generalised predictive controller (GPC) four parameters have to be specified. To choose those parameters is not a trivial task since they are not directly related to control or regulation performance. The presented tuning method exploits model-parameters to select suitable controller parameters. Additionally, a Rhinehart filter is incorporated in the design to decrease the impact of noise, therefore, a fifth parameter has to be optimised. The proposed method has been tested in simulation and on a real system.