Linking Nonlinear Steady-state and Target Set-point Optimisation for Model Predictive Control

The problem of cooperation of model predictive control (MPC) algorithm with nonlinear (local) steady-state optimisation is considered. It is important when dynamics of disturbances is comparable with dynamics of the process, thus when application of a classical multi-layer approach may be not efficient. The case concerns a situation when a reliable steady-state process model is nonlinear, often difficult for optimisation and standard approach of linear target set-point recalculation combined with the MPC algorithm does not suffice. Nonlinear formulations of the target set-point optimisation (integrated with the MPC, performed at every sampling instant) are proposed in the paper. The best approach, but rarely implementable in real-time, would be to apply full nonlinear optimisation. Approximate, linear-quadratic or piecewise-linear formulations of the target set-point optimisation are proposed, depending on problem properties. The approach is illustrated on a distillation column example. Copyright © 2006 USTARTH