Refrigeration is essential to many sectors of society, including food preservation, the pharmaceutical industry, industrial processes, and thermal comfort. Current refrigeration systems, based on vapor compression, must become more energy-efficient and reduce their environmental impact. However, testing different low global warming potential refrigerants requires time and test benches. Several mathematical or heuristic models can simulate the behavior of a refrigeration system. These models allow computationally testing different types of experimental benches or configurations. This study utilizes eXtreme Gradient Boosting, a machine learning algorithm, to simulate the performance of a refrigeration system using a minimal amount of experimental data. This model uses temperatures and secondary flows in the condenser and evaporator as inputs. With this information, the model can predict the behavior of the coefficient of performance, power consumption, cooling capacity, refrigerant flow rate, and up to four refrigerants, such as R134a, R513A, R516A, and R1234ze(E). Additionally, the model achieves an overall coefficient of determination greater than 0.98 and an overall accuracy of 96.13 %. These enable the simulation of different operating values of the experimental bench to determine the most effective refrigerant for achieving the highest coefficient of performance; thus, R1234ze(E) and R513A are those that represent the highest COP.
扫码关注我们
求助内容:
应助结果提醒方式:
