An ADMM algorithm for incorporating structural constraints in self-optimizing control

Abstract

An ADMM algorithm is proposed for selecting structurally constrained measurement combinations as controlled variables (CVs). The CV selection is based on the self-optimizing control principle, where the goal is to choose CVs such that the steady-state operation is optimized when they are kept at constant set-point. When CV selection incorporates structural constraints, it becomes a non-convex optimization problem and thus, finding the optimal solution is difficult. However, using an ADMM algorithm for a given measurement set together with specified structural constraints, a local solution can be obtained. The resulting CVs seem to give similar or better performance when compared to other existing methods. The proposed method was evaluated on case studies, consisting of a binary distillation column and an evaporator.

An ADMM algorithm for incorporating structural constraints in self-optimizing control