Design and Optimization of an S-CO₂ Turbine Based on Gauss Process Regression

To solve the problem of poor accuracy and long period of design and optimization for a supercritical carbon dioxide (S-CO₂) turbine using traditional design method, a fast thermodynamic design method was established for the radial inflow turbine based on one-dimensional flow theory. A design-optimization method was also proposed based on Gauss process regression, which combines the high precision three-dimensional aerodynamic analysis with the thermodynamic design to evaluate the real efficiency of the turbine aerodynamic design, verify the accuracy of the design results in simulated annealing process, and to demonstrate the effectiveness of the design and optimization for the S-CO₂ turbine with calculation examples. Results show that via the method proposed, the isentropic efficiency of the turbine could be increased from the original 83.68% to the optimal 91.20%; an optimal design result would require 120 cycles of aerodynamic analysis by traditional method based on simulated annealing optimization, however, by the method proposed, only 24 cycles are requred, which greatly shortens the time of design and optimization, and therefore may serve as a reference in engineering applications.