Modeling of thermophysical properties of refrigerants using neural networks for improving heat pump optimization.

Heat pump systems (HPSs) compete with low-cost heating technologies like gas-fired boilers mainly because of their role in decarbonizing the building sector. Considering system dynamics inside the HPS results in many interdependencies between the HP components and the refrigerant selection, and thus, in large-scale models that are hard to optimize. Therefore, this work proposes a method that enhances the simultaneous optimization of the design and operation of electrically driven HPSs. The developed method uses artificial neural networks to surrogate the calculations of the thermophysical properties of refrigerants. Five fluids were evaluated to determine the next-generation refrigerants based on thermodynamic, economic, and environmental criteria. Deterministic global optimization results in a significant decrease in the HP capacity of fluoroethane and propane to 4.23 kW and 4.95 kW, compared to the normative design. The savings in total annual costs reach 51% and 45% in the case of difluoroethane and isobutane or propane, respectively.