Comparison of a “model of the model” approach with directly measured reference values for assessing the energy efficiency of a low-temperature refrigerating system

This paper is addressed at owners / operators of direct refrigerating systems (without additional water / glycol loop) who are interested in monitoring the energy efficiency of their systems but do not have the possibility to measure refrigerant mass flow directly in their systems.

Specification VDMA 24247-7 (2021) deals with energy efficiency of refrigerating systems. In this specification, several so-called indirect measuring methods are suggested as an alternative to a direct measurement of refrigerant mass flow and thus cooling capacity of the system. One of these indirect measuring methods is the compressor model. This method is based on data of the manufacturer of the compressor gained under test stand conditions in a laboratory. In previous publications, it could be shown that the approach recommended in VDMA 24247-7 shows a very good agreement with directly measured reference values for various operating conditions using monitoring data of a real-world low-temperature refrigerating system of a cafeteria. However, as re-occurring operating conditions are calculated repeatedly offline in a 32-bit spreadsheet application, computational effort of the method can be called unnecessary high at least and handling of the method awkward at best. Thus, the presently available method delivers good results in a stand-alone analysis of previously collected data but it is not appropriate for a real-time online monitoring of energy efficiency of refrigerating systems. In order to remedy these shortcomings, the authors developed an alternative based on a “model of the model” approach.

For this approach, a complete performance map of the compressor installed in the aforementioned low-temperature refrigerating system was calculated using the compressor model. Output variables comprise cooling capacity, electrical power consumption of the compressor, refrigerant mass flow as well as Coefficient of Performance. Next, a regression model of the behaviour of the refrigerating system was developed using the performance map data in a supervised learning domain. In a final step, the results from the newly developed “model of the model” are compared to results from the original compressor model as well as to directly measured values, which serve as a reference. The comparison is done for different scenarios – including low, medium and high refrigerant mass flow as well as quasi-constant and intermittent operating conditions.