Conference Year



Distributed, ESC, VRF


The variable refrigerant flow (VRF) technology has facilitated the development of multi-split ductless air conditioning systems, in which multiple indoor units (IDU) are used to regulate the refrigerant flow to achieve individualized zoning control. Model based control for VRF system demands for more modeling efforts in part due to diverse configuration, as well as changes in load and ambient conditions. As a model-free control strategy, Extremum Seeking Control (ESC) has been investigated for VRF systems. Dong et al. (2015) applied the standard centralized ESC scheme to a VRF system that consists of one outdoor unit (ODU) and four IDU’s. Simulation results have indicated the effectiveness of such strategy. As the number of IDU’s increases, the complexity of centralized controllers will increase accordingly. Therefore distributed ESC becomes a natural consideration for VRF systems with large number of IDU’s. In this paper, the Shashahani gradient based distributed ESC scheme proposed by Poveda and Quijano (2013, 2015), is applied to the four-zone VRF system simulated by Dong et al. (2015). In particular, this scheme is enhanced by appending a band-pass filter array at the output to achieve a better “isolation” among individual input channels. A single-input ESC is applied to the ODU, while the distributed ESC is applied to the four IDU’s with each acting as an agent. For each agent, the respective power consumption is used as feedback. The objective is to minimize the total power consumption of all agents. For the ODU ESC, the compressor suction pressure (PCS) set-point is employed as the manipulative input. For the IDU DESC, the evaporator superheat (SH) set-point is used as the manipulative input for each IDU agent. The distributed ESC scheme assumes full information communication among all IDU’s. Simulation study is performed to evaluate the proposed strategy with the Modelica based dynamic simulation model developed by Dong et al. (2015). The ESC is designed under the ambient condition of 35oC and 40 %RH, respectively. The initial temperature of all four IDUs zone is 29oC, and the zone temperature set-point is 26oC. The heat loads for IDU1 through IDU4 are 3000W, 2600W, 2400W and 2000W, respectively. It takes the average total power about 10000 seconds to converge to about 3200W in steady state, with PCS around 13bar, and the SH values of IDU1 through IDU4 at 4.5oC, 4.5oC, 6oC, and 5.5oC, respectively. The total power consumption was decreased from 4500 W to 3200 W, i.e. by 29%. In comparison with the centralized ESC Dong et al. (2015), the steady state error of total power is less than 50w. Work is under way to improve transient and steady-state performance, as well as simulation of other operation modes.   Â