Application of artificial neural networks for generation of energetic maps of a variable speed compression system working with R1234yf

This paper proposes a new tool that uses artificial neural networks to build energetic maps for a vapor compression system working with R1234yf. From these energetic maps, it is possible to visualize and identify the zones with the best performance. Additionally, it was concluded that the temperature of the condensing agent and the brine has a greater influence on the COP than their volumetric flows. l computer simulations were performed to analyze the impact of changing the configuration of the artificial neural network. A hybrid method to train the artificial neural network was used; this method was a combination of: simulated annealing, regression, and the conjugate gradient in multi-dimensions. sults show that artificial neural networks can be used to predict the COP of a vapor compression system, and to analyze how several input parameters of the system may affect its energetic performance. Additionally and for energetic comparison purposes, one artificial neural network was trained using data from a compression system operating with R134a, and another artificial neural network was trained with R1234yf.