DESIGN AND EXPERIMENTAL VALIDATION OF SUBSPACE PREDICTIVE CONTROL

Abstract

This Thesis studies different subspace identification techniques and its corporation with predictive control to introduce a subspace predictive controller (SPC). Furthermore, it investigates the combination of this controller with adaptation algorithm to develop an Adaptive version of Subspace Predictive Control which has the ability to deal with time varying and nonlinear processes. SPC is proved as a competitive replacement for conventional MPC as it avoids system’s states measurement and hence avoids the need for state observer and its design. Instead, SPC predicts the output based on a bank of collected input and output data in a manner similar to Generalized Predictive Control (GPC) but can be extended to deal with MIMO systems based on a nonparametric model. Finally, the algorithm has been validated experimentally and applied to two benchmark systems Servomechanism and Level Control Process. The obtained results are promising.