| This thesis addresses the development of the Reference System Model Predictive Control (RS-MPC) algorithm for nonlinear processes operating over a wide range of conditions. The proposed formulations are shown to be an effective approach for addressing the transition control problem of highly nonlinear processes. This includes systems with changes in the magnitude and sign of the steady state gains, dead times, and non-minimum phase characteristics.; The proposed RS-MPC introduces to the control design a set of requirements, which set the desired closed-loop behavior in reference to an achievable performance. The main characteristic of this approach is the fact that the controller design explicitly uses information about the structural dynamic limitations of the process. Consequently, it does not force the controller to requirements that are known to be impossible to achieve. Furthermore, during transitions, the algorithm does not use a fixed reference system. The reference system is generated on-line by linearizing the plant model and it is based on local invertibility constraints.; The technique is illustrated using constrained setpoint tracking and disturbance rejection of highly nonlinear chemical reactors. It has been shown that the proposed controller deals satisfactorily with many kinds of operating scenarios. It is concluded that for broad transition problems, the proposed RS-MPC is an excellent alternative to conventional MPC. |
