Fast Calculation, Thermodynamic properties, Interpolation
It has been previously demonstrated that the most significant computational requirements for vapor compression system models are associated with evaluation of thermodynamic properties. This is particularly important for transient models because properties are evaluated at each time step and the overall model can often run less than real time. The typical approach for evaluating thermodynamic properties involves the use of complicated equations of state (EOS), such as utilized in standard software tools such as RefProp and CoolProp. Overall computation speed can be significantly enhanced using interpolation methods that are based on pre-calculated thermodynamic properties. This paper presents an improved interpolation method to quickly and accurately retrieve refrigerant properties based on the neural networks. Since the approach has an explicit functional form, it is able to avoid the computation time to find nearest points in the thermodynamic database. Comparisons between the proposed method and Refprop are provided for a wide range of pressure and enthalpy to show its accuracy. Then, performance comparisons between the proposed and conventional interpolation methods for a transient vapor compression cycle simulation are provided.