Model Predictive Control Using Approximate Models

In light of its successful application to the control of SISO processes, the principles of Wiener MPC are extended to the MIMO case and implemented on a pilot scale ethanol/water distillation column. Wiener models consist of a linear dynamic element with a static non linear transformation on the output and therefore allow the introduction of some degree of non-linearity to a system representation without the added complexity associated with non linear models. In addition to this, they may be incorporated into the MPC structure in such a way that the nonlinearily is essentially removed from the optimisation problem, resulting in a simple linear optimisation as opposed to a complicated nonlinear one. The implementation of the MIMO Wiener MPC is performed using an ABB MOD.WO distributed control system and a VAX workstation and is shown to require little more effort to apply than linear MPC.