Distributed predictive control for complex hybrid system. The Refrigeration System example

In this paper, a distributed model predictive control (DMPC) is applied to the supermarket refrigeration system benchmark specified by the HYCON project. This system is an example of continuous nonlinear system controlled by on/off variables. The aim of the proposed approach is to overcome the computational complexity of previous centralized MPC by distributing the control while keeping the communication among controllers at a low level. Two types of interactions among controllers are considered: for some variable the predicted values are broadcast to involved controller while for others a so-called price mechanism is used.