A computationally efficient nonlinear MPC algorithm

In this paper, a novel Model Predictive Control (MPC) algorithm for control of nonlinear multivariable systems is proposed. The on-line computational demand of the algorithm depends only on the number of manipulated variables; it does not depend on the input (or control) horizon. Thus, the on-line computational demand is significantly smaller than conventional nonlinear Model Predictive Control algorithms which attempt to solve the on-line optimization problems exactly. We show that asymptotic stability can be guaranteed in some cases. Its feasibility for practical implementation is demonstrated on a distillation column dual composition control problem using a rigorous tray-by-tray model (with input horizon of 10).