Truncated step response models for model predictive control

Abstract

A method for deriving truncated step response models for model predictive control (MPC) is derived and demonstrated. Truncated step response models can result in a dramatic reduction in the number of step response coefficients needed, and can therefore significantly reduce the computational load on the control system. The truncated models are derived in a way which attempts to minimize the robustness degradation caused by the modelling error introduced by truncation. The truncation of the step response model only affects the state estimation part of the MPC. The algorithm for finding the control move is not affected as long as the prediction horizon is kept smaller than the truncation time. © 1993.