Performance improvement and multi-objective optimization of energy, economic, and environmental factors in organic Rankine cycle using machine learning-driven quasi-two-stage single screw expander

This study introduces a quasi-two-stage single screw expander (QTS-SSE) designed to improve the performance of organic Rankine cycle (ORC) systems. A predictive model specifically for the QTS-SSE was developed using machine learning techniques. Thermodynamic, economic, and environmental models were developed to analyze the impact of key adjustable devices, including the QTS-SSE, working fluid pump, and cooling water pump, on net power (Pnet), net efficiency (ηORC), electricity production cost (EPC), and equivalent CO2 emissions (ECE). Results show that the prediction error of QTS-SSE model is minimal. The velocity expansion power reaches 0.56 kW, which enhances the isentropic efficiency of expander by 2.48 %. Both Pnet and ηORC increased by 5.31 % each, while EPC and ECE decreased by 4.25 % and 5.14 %, respectively. The maximum efficiency of the working fluid pump was 30 %, while the water pump was 62.6 %. Additionally, multi-objective particle swarm optimization was employed to balance energy, economic, and environmental performance. The innovations in this paper enrich the theoretical frameworks of volumetric expanders, broaden efficient operating conditions, explore mechanisms by which the novel expander improves ORC systems, and provide a reliable reference for researchers to optimize overall ORC system performance.