Closed-loop model predictive control

Project members: D.H. van Hessem, O.H. Bosgra

Sponsored by: GROWTH Program of the European Union

During the last twenty years, process control and optimization has focused on open-loop predictive methods. The main disadvantage of these open-loop methods is that feedback is only generated by means of a receding horizon strategy. Consequently, model predictive control suffers from the limitation of any open-loop strategy, namely that the possibility of shaping the process sensitivity, a basic characteristic of feedback design methods, is completely absent. As a consequence, robustness is and always has been a problem with MPC. To overcome these inconsistencies a closed-loop formulation has been developed making an explicit use of both feedforward and feedback in a generalized plant setup. An example of such a control solution is visualized in Figure 13.

Figure 13: Visualization of a closed-loop predictive control solution.