Optimization of organic Rankine cycle turbine expander based on radial basis function neural network and nondominated sorting genetic algorithm II

The organic Rankine cycle (ORC) represents an effective technology for the recovery of medium- and low-temperature waste heat. Within this system, the turbine expander plays a critical role in determining the reliability and efficiency of the overall process. This paper presents a structural optimization approach that integrates a radial basis function (RBF) neural network model with the nondominated sorting genetic algorithm II (NSGA-II), considering the isentropic efficiency and power of the ORC turbine expander using R1233zd(E) as the optimization objectives. Utilizing the design-of-experiments method in conjunction with simulation, a high-precision RBF neural network model was developed and trained. The external performance and internal flow characteristics of the original and optimized model are compared. In addition, the entropy production method is used to locate and quantitatively evaluate the energy losses. The results indicate that the RBF neural network model exhibits high predictive accuracy, with a correlation coefficient (R2) exceeding 0.9 for both objective functions. The optimization process significantly enhanced the performance of the ORC turbine expander. Under Q/Qd = 1.2, the isentropic efficiency and power are significantly improved by 6.13% and 33.96%. The optimized model can accommodate a larger range of flow variations, increasing the efficient operation region by 1.28 times. The energy loss of the ORC turbine expander decreases by an average of over 17% due to the effective suppression of vortices at the leading edge and outlet of the impeller. This work provides a valuable reference for improving the performance of radial turbine expanders for waste heat recovery and other application fields.