Adaptive multivariable optimal predictive control

Abstract

Abstract A multivariable control strategy that incorporates long‐term prediction is presented. The future control policy is calculated so as to minimize a weighted sum of squared input and output deviations from reference values. Constraints on inputs are explicitly incorporated in the minimization to allow optimum operation of the process. The control approach is developed to allow either adaptive operation or operation using a constant model that has been identified on‐ or off‐line. Several simulation examples are presented to demonstrate the robustness of the controller to unknown load dusturbances, process‐model mismatch, time‐varying dynamics and non‐minimum phase characteristics.