Model Predictive Control of Variable Refrigerant Flow Systems for Room Temperature Control

This paper presents a supervisor model-switch method for model predictive control of room temperature using a one-to-three variable refrigerant flow (VRF) system. This method utilizes system identification to obtain linearized transfer functions of the variable refrigerant flow system’s compressor speed and electronic expansion valve to room and superheat temperature. The transfer functions for room and the two-phase region temperatures of the evaporator are derived from the energy conservation law to apply in VRF systems. The model predictive control structure is adopted to consider the coupling effects between control states. The effects of superheat and room temperature error weighting factors on rapid cooling and steady-state error are numerically investigated. In addition, this research utilizes a switching linearization model approach to overcome the limitation of a single identification model with a narrow, accurate temperature control range. A supervisor model-switch strategy is proposed to switch multiple models for the same target temperature to average the risk of model mismatch, and calculate the minimum allowable switching time to ensure system stability during switching, thereby expanding the range of temperature control. The proposed method has been experimentally evaluated on a one-to-three VRF system to control the steady-state temperature of each indoor unit within 0.5°C and maintain a superheat above 0°C to avoid liquid refrigerant from entering the compressor.