Machine learning models for high-accuracy energy and exergy prediction of low-GWP R134a/R1234yf blends
摘要
A mathematical model of the vapor compression refrigeration cycle for different operating conditions (different evaporator and condenser temperatures, different superheat and supercooling temperatures) is established in the present study. Instead of R134a, mixtures of R134a and R1234yf with lower global warming potential values are evaluated as the working fluid in the vapor compression refrigeration cycle. The energy and exergy performances of these refrigerant mixtures were estimated by machine learning algorithms. Seven different machine learning algorithms have been used. These are support vector regression, random forest, extreme gradient boosting regressor (XGBR), CatBoost, light gradient boosting machine, adaptive boosting, and decision tree. The XGBR algorithm provided higher accuracy in the energy efficiency results of both refrigerants, with the best performance of the CatBoost algorithm in terms of exergy efficiency. For the blend of R134a/R1234yf (10/90), taking the evaporator temperatures as − 25 °C, − 20 °C, − 15 °C, − 10 °C, − 5 °C, and 0 °C, the condenser temperature as 35 °C and both superheating and subcooling temperatures as 7 °C, COP values vary between 2.6723 and 5.7909 and