Evaluation of a virtual refrigerant charge sensor

Studies have shown that more than 50 percent of packaged air conditioning systems are improperly charged. Improper refrigerant charge can increase energy usage, reduce capacity, and decrease equipment lifespan. Furthermore, refrigerant charge leakage can contribute to global warming in the long term. The primary goal of the thesis is to evaluate and enhance the virtual refrigerant charge sensor, developed by Li and Braun (2006a). The virtual refrigerant charge sensor algorithm employs low-cost and non-invasive measurements (i.e. surface mounted temperature measurements) to estimate refrigerant charge level for packaged air conditioning systems. It can be embedded within a portable device (i.e. a PDA) for a technician\u27s use in the field or permanently installed on units. To evaluate the accuracy of the virtual refrigerant charge sensor, data collected from previous laboratory tests and a manufacturer were utilized. In addition, new laboratory tests and simulations were performed to obtain data for various system types over a wider range of operating conditions. The systems for the new laboratory tests included residential split systems that employ fixed capillary tubes, or thermostatic expansion valves and R-22 or R-410a as the refrigerant. To assess the impact of refrigerant charge level on system performance, capacity, efficiency, and operating cost were calculated from the data for the units tested in the laboratory at all test conditions and for some of the data obtained from the manufacturer. Based on the evaluations, the virtual charge sensor was found work well in estimating the refrigerant charge for systems that do not utilize accumulators when using the original default parameters. For systems with accumulators, however, the parameters needed to be improved. A new method for determining default parameters was developed that depends on three elements: liquid line length, rated subcooling, and rated charge. The liquid line length is particularly important because a substantial amount of refrigerant is stored as liquid. The parameters decreased the errors between the actual and predicted charge. Even better performance was achieved for the virtual refrigerant charge sensor when the improved parameters were tuned, minimizing the errors by using test data and linear regression. Overall, the enhanced method provided estimates of refrigerant charge that were within 10% of the actual charge over a wide range of operating conditions for a number of different systems