Extending the virtual refrigerant charge sensor (VRC) for variable refrigerant flow (VRF) air conditioning system using data-based analysis methods. (25th January 2016)

• Record Type: Journal Article
• Authors: Li, Guannan; Hu, Yunpeng; Chen, Huanxin; Shen, Limei; Li, Haorong; Li, Jiong; Hu, Wenju
• Is Part Of: Applied thermal engineering. Volume 93(2016:Jan.)
• Page Start: 908
• Page End: 919
• Publication Date: 2016-01-25
• Subjects: Accumulator -- Correlation coefficient analysis -- Refrigerant charge amount -- Support vector regression -- Variable refrigerant flow -- Virtual refrigerant charge sensor; Heat engineering -- Periodicals; Heating -- Equipment and supplies -- Periodicals; Periodicals; 621.40205
• DOI: 10.1016/j.applthermaleng.2015.10.050
• Languages: English
• ISSNs: 1359-4311

Abstract

Highlights: VRC models can only predict VRF system charge amount at undercharge situation. VRC models show poor performance for refrigerant charge diagnostic purposes. VRC models need modifications when charge levels are larger than 90%. Correlation analysis proved accumulator associated with VRC estimate errors most. Support vector regression reduced the VRC errors at overcharge situations. Abstract: A proper refrigerant charge amount (RCA) prediction algorithm is important to air conditioning systems. In variable refrigerant flow (VRF) systems, the traditional virtual refrigerant charge (VRC) sensor models perform well at undercharge situations but produce large prediction errors at overcharge situations. When the refrigerant charge level (RCL) is over 90%, the correlation coefficient data-based method was introduced to select the additional input variables to modify the VRC models. Two data-based algorithms, multiple linear regression (MLR) and non-linear support vector regression (SVR), were used to improve the prediction performance. The prediction performance of the pure SVR model was also compared. Results reveal that the overall prediction errors for SVR based modified VRC model (SVR-VRC) is 5.53%, the minimum among the four models. The SVR-VRC model improves the VRC models and extends the application in the VRF system when only the system self-provided sensor measurements are used.