Conference Year



Multiple Outdoor Unit Variable Refrigerant Flow System, Extremum Seeking Control, Energy Efficiency, Real-time Optimization, Automatic Staging


For efficient operation of a variable refrigerant flow (VRF) air conditioning system with multiple outdoor units (ODUs), we propose a model-free control strategy based on extremum seeking control along with automatic staging control logic. The proposed strategy is evaluated with a representative VRF system consisting of 12 indoor units (IDUs) and three ODUs. The IDU zone temperature is regulated by EEV opening, and the compressor pressure is regulated by compressor speed. To optimize load sharing among multiple ODUs in operation, a set of bypass valves (BPVs) are added to the suction side of the compressors to manipulate refrigerant flow distribution among different compressors as needed. A penalty-function based multivariable extremum seeking control (ESC) method is used for real-time optimization of system operation. The performance index as the ESC feedback is the total power of the compressors, the ODU fans and the IDU fans, augmented with penalties for securing minimum superheat at the suction side of compressors. The manipulated inputs include the compressor suction pressure setpoint, the openings of BPVs at the suction side of the compressors, and a uniform setpoint of fan speed for all ODUs. As for the ESC feedback, the compressor power is normalized by its capacity. A set of control strategies for staging on/off particular ODUs is developed based on the compressor speed of the operating ODUs. Under increasing load, if the operating compressor(s) speed exceeds the higher limit of operation speed range (80% of rated speed), an additional ODU turned on to meet the load demand. Under decreasing load, it is desirable to turn off the least efficient ODU in a model-free fashion. In this study, an ESC based ODU staging-off strategy is proposed, for which the compressor shaft power normalized by the rated capacity is adopted as the ESC input. In addition to the compressor pressure setpoints and ODU fan speeds, the manipulated inputs of ESC also include the openings of suction-side BPVs in order to optimize load sharing among the multiple ODUs. With online optimization of ODU load sharing based on the normalized compressor power, the ESC can drive less efficient compressor(s) to operate at lower speed/capacity. If the compressor speed of an ODU falls below the preset lower limit of operational speed range (e.g. 20% of the rated speed) for long enough time, this ODU will be turned off. A dynamic simulation model of the multi-ODU VRF system is developed with Dymola and TIL Library. Simulation studies have been performed to evaluate the proposed ESC strategy for energy efficient operation during constant load patterns and the control logic for staging on and off ODU during load increase and decrease. The total power searched by the ESC is shown to be close to that obtained by a genetic algorithm based global optimization procedure in Dymola. Also, ESC is shown to be able to turn off least efficient ODU during load decrease without model knowledge. The load-sharing BPV at the compressor suction-side demonstrates bearable pressure loss except for the scenarios of large split ratio.