How to evaluate the performance of sub-critical Organic Rankine Cycle from key properties of working fluids by group contribution methods?

An artificial neural network (ANN) model is developed to predict the ORC performance from key properties of working fluids, including critical temperature, critical pressure, acentric factor and ideal gas heat capacity, based on the 5400 calculated data from REFPROP for 54 working fluids. When these key properties are unknown for working fluids, group contribution methods (GCMs) are employed to combine with the established ANN. For the considered three GCM-ANN models, 21 potential working fluids of ORC are used to evaluate the accuracy in the prediction of key properties and cycle parameters. From the obtained results, it can be concluded that the developed ANN has average absolute deviations (AADs) 5.9866%, 0.1024%, 0.9684%, 0.1131% and 1.6283% for pump work, evaporation heat, turbine work, condensation heat and cycle efficiency, respectively. For key properties, accuracy of critical temperature has the most significant effect on the ORC predictions. As for the three GCMs, SU-GCM has the least deviations for the prediction of properties. The corresponding AADs of GCM-ANN model are 15.89%, 10.73%, 12.88%, 10.40% and 2.51% for pump work, evaporation heat, turbine work, condensation heat and cycle efficiency, respectively. When key properties are obtained from experiments or GCMs, the developed ANN can be applied to predict ORC performances for any working fluid easily and quickly.