A review of recent advancements of variable refrigerant flow air-conditioning systems

• Published in: Applied thermal engineering 2020-03, Vol.169, p.114893, Article 114893
• Main Authors: Wan, Hanlong, Cao, Tao, Hwang, Yunho, Oh, Saikee
• Format: Article
• Language: English
• Subjects: Air conditioning; Algorithms; Coefficient of performance; Computer architecture; Computer simulation; Control methods; Data analysis; Energy storage; Fault detection; Fault diagnosis; Fault tolerance; Literature review; Machine learning; Model accuracy; Modelling; Refrigerants; Refrigeration; Ventilation; VRF system
• ISSN: 1359-4311; 1873-5606
• DOI: 10.1016/j.applthermaleng.2019.114893

•Many VRF studies focused on providing ventilation and temperature humidity independent control function.•VRF models are either in steady-state and transient state groups or empirical and component-based groups.•Machine learning and data mining technologies are used in VRF modeling, control, and FDD studies.•Reversed cycle defrost is the most common method for VRF system defrost. Variable refrigerant flow air-conditioning (VRF) systems are important and widely used building energy systems around the world. This study reviews recent developments of the VRF systems in system architecture development, modeling, experiment, control strategies, fault-detection-and-diagnosis, and defrost. Except for the defrost study, each section is classified according to the research targets or methods. The strengths, drawbacks, challenges of current studies and possible solutions are discussed for each section. Then, since the modeling, simulation, control, and fault detection all required data analysis, a separate section was used to summarize. In conclusion, researchers have added new functional devices like outdoor air process ventilation to address the lack of ventilation functions, developed novel VRF system resulting in up to 45% increase in the coefficient of performance compared to previous experiments, and improved the model accuracy within 15% agreement. Data analysis methods include conventional methods and knowledge-based methods. The conventional method can explain the system clearly but lack robust. The knowledge-based method can be used easily but hard to explain. Future development could be made on an integrated VRF and energy storage system to provide higher flexibility or on a new algorithm that can combine the benefits of both data-driven methods and traditional component-based methods for modeling and control.