Long-Range Predictive Control Using Weighting-Sequence Models

摘要

The predictive control algorithm (PCA), dynamic matrix control (DMC) and identification/command (IDCOM) process control algorithms are compared. It is shown that they are capable of providing good stable control under certain conditions. The IDCOM, however, is seen by simulations to be the least satisfactory approach. The derivations show that all the methods are capable of further refinement; in particular DMC (potentially the most robust) benefits from using a proper predictor and cost-function rather than the heuristics of the original approach. The applicability of the algorithms and the restrictions imposed by the simple assumed plant model are discussed.