An analytic price-driven coordination scheme for distributed model predictive control systems

In this work, we propose an analytic approach to coordinate distributed model predictive control (MPC) systems that are subject to equality constraints. The main idea behind the coordinated distributed model predictive (CDMPC) is to improve the performance of an existing decentralized control system with a minor modification applied to the existing distributed control system. As such, the enhancement can be interpreted as a bilevel optimization problem, which consists of modified local controllers and a coordination level that ensures optimal centralized behavior of the plant. The modification to the decentralized MPC controllers would be equivalent to relaxing local versions of the overall interaction constraint using a price vector to penalize violations. In the proposed price-driven CDMPC scheme, local variables are parameterized in terms of the price vector. As a result, an analytic closed-loop solution to the optimal price vector can be developed. The effectiveness of the proposed CDMPC is illustrated using a chemical process example.