Computational issues in nonlinear predictive control

Abstract

This paper discusses the implementation of different nonlinear strategies in a model predictive control (MPC) framework to control an exothermic continuous stirred tank reactor. The computational efficiency of an MPC strategy depends on the method used to predict model outputs within the optimization loop. The computational requirements of collocation and numerical-based methods to solve nonlinear differential modeling equations are compared with the nonlinear quadratic dynamic matrix control (NLQDMC) formulation. The convolution coefficients for NLQDMC are obtained using analytical and numerical methods and their computational time requirements are compared.