Pre-Diploma Thesis Robust Switching Strategies in Self-Optimizing Tracking Control

Self-optimizing control is when acceptable performance can be achieved using pre-computed references and with no need to re-optimize when disturbances occur. For linear steady state processes, several methods to identify selfoptimizing control structures have been developed recently. For many industries, however, unsteady state processes become more and more important. A typical example is batch distillation. This thesis extends the idea of self-optimizing control to such tracking problems. The main di erence is that the optimal input trajectories in batch processes often contain a nite number of jump discontinuities. These divide the batch into several intervals and for each interval exists one self-optimizing controller. It is now very important to activate the next controller at the right time as failure to do so may lead to infeasibility. Hence, good feedback-based switching laws are desirable. A promising candidate is the so-called nullspace method which is examined rst. It is found, that a linear combination of states does not lead to acceptable results. Therefore, an alternative switching strategy that uses indirect information about the disturbance is presented. In combination with simple PI-type continuous controllers, self-optimizing control for this particular nonlinear batch process becomes possible and leads to good results.