Evaluation of the energy performance of variable refrigerant flow systems using dynamic energy benchmarks based on data mining techniques

The variable refrigerant flow (VRF) system has extremely different energy performance at various operation conditions. Its power consumption is inconsistent even under the steady operation condition. In order to accurately evaluate the VRF system’s dynamic energy performance, this study proposed a data-mining-based method to benchmark and assess its energy uses. The correlation analysis is used for key factors selection and the interquartile range rule is employed to remove outliers of the database. In addition, the power consumption patterns are classified using decision tree (DT) method. The classification results are validated by the ANOVA analysis and post hoc test. Nine energy benchmarks are established based on the classified power consumption patterns. Moreover, an energy consumption rating system is established to provide quantitative assessment on the power consumption of the VRF system. A case study is conducted by comparatively analyzing the energy performance of the VRF system at multiple refrigerant charge fault cases. Results show that both the PLR and OT significantly affected the power consumption of the VRF system. However, the degree to which the refrigerant charge fault affects system power consumption varies with the power consumption patterns. For different patterns, the power consumptions of the VRF system were either lower, higher or similar to each other at various RCLs. Results also suggest that the energy benchmarking process provide reasonable classification criteria, and the grading process provide quantitative assessment on the energy consumption. Therefore, the proposed dynamic energy benchmarks are reliable and reasonable to evaluate the dynamic energy performance of VRF systems.