(311b) Reverse Strategy for the Molecule-to-Flowsheet Design of Hen-ORC System: Screening for Novel HFO Working Fluids and Process Structures

The utilization of waste heat is crucial for mitigating industrial carbon emissions. Among the promising methods, the Organic Rankine Cycle (ORC) has garnered significant attention for its capacity to generate electricity from low-temperature heat sources ranging from 100 to 300°C. Many factors affect the thermodynamic performance of ORCs, such as the temperature level of waste heat in the heat exchanger network (HEN), the selection of working fluids, and ORC operating conditions. In recent decades, the integrated design of HEN-ORC has been a dynamic field of research, while the integrated design of HEN-ORC and new working fluids of hydrofluoroolefins (HFOs) remain relatively limited. It is a complex optimization problem with a huge design space combining continuous operating conditions with discrete HEN-ORC configurations and molecule space. Besides, there is a significant dearth of thermodynamic data on new working fluids in existing literature and databases. For modeling the thermodynamic behaviors of new working fluids in ORCs, the incorporation of the equation of state (EOS) or data-driven thermodynamic property prediction models introduces additional non-linearities.

To overcome this difficulty, a reverse design strategy from the HEN-ORC system to the HFO working fluid is proposed. In the proposed strategy, a quantitative EOS-free heat-to-work model of ORC is developed, thus avoiding complex non-linearities from thermodynamic property calculation. The optimal HEN-ORC structure and the thermodynamic properties of the hypothetical working fluid can be obtained by an EOS-free HEN-ORC model, which is a mixed integer linear programming (MINLP) problem. To find the best HFO candidate, whose properties match the hypothetical working fluid, two GC-ANN models are developed to predict the thermodynamic properties of the novel HFO candidates. One HFO working fluid database containing more than 430,000 hydrofluoroolefins (HFOs) is established as the HFO candidate pool, and one hydrocarbon working fluid database is constructed to provide physical property data for training GC-ANN models. The presented method is employed in two HEN-ORC cases, where new working fluids are found. The total annual cost of Case 1 is 12%~22% lower than the literature, and the work output of Case 2 is 5~8% higher than the literature.