Offset-free model predictive control of a vapor compression cycle

This work describes the application of an offset-free model predictive controller (OF-MPC) to a vapor compression cycle (VCC). To this end, first a linear model is identified from data generated from a first-principle model of a VCC, interfaced with a building model implemented in EnergyPlus. Next, a model predictive controller is designed that includes an augmented model (including disturbance states) and an associated Luenberger observer to estimate the disturbance (plant-model mismatch) at steady state and eliminate it. Tuning guidelines are presented that enable partial ‘decoupling’ of the Luenberger design and choice of MPC parameters, leading to offset elimination and superior tracking performance while reducing the energy demand on the VCC, relative to traditional control approaches. The superior closed-loop performance of the OF-MPC strategy is demonstrated for a VCC model in isolation, subject to realistic disturbances and subject to measurement noise.