A reconfigurable Plug-and-Play Model Predictive Controller for multi-evaporator vapor compression systems

Abstract

This paper presents a reconfigurable Plug-and-Play (PnP) Model Predictive Controller (MPC) for multi-evaporator vapor compression systems (VCS) where individual evaporators are permitted to turn on or off. This alters the number of performance variables, actuators and constraints. The proposed approach features structural online updates of the closed loop system with stability guarantees, and avoids the need to commission and tune separate controllers for when individual subsystems are turned on or off. To compare the performance of the proposed approach, a more conventional switched MPC is also developed in order to provide a benchmark design, wherein separate model representations are developed and controllers with numerous tuning parameters are synthesized and deployed depending on the VCS operation mode. Simulations are provided comparing the performance of the proposed reconfigurable PnP MPC to the traditionally-designed switched MPC. Results confirm that the reconfigurable PnP MPC maintains the same performance as the switched MPC approach in terms of room temperature reference tracking after zones are switched on, enforcement of critical machine constraints, and disturbance rejection. However, reconfigurable PnP MPC requires no extra tuning or controller design effort, and can be automatically synthesized from a single master controller design for any VCS operating mode.