A hybrid approach to optimization of radial inflow turbine with principal component analysis

Energy conversion efficiency is one of the most important features of power systems as it greatly influences the economic balance. The efficiency can be increased in many ways. One of them is to optimize individual components of the power plant. In most Organic Rankine Cycle (ORC) systems the power is created in the turbine and these systems can benefit from effective turbine optimization. The paper presents the use of two kinds of hybrid stochastic/deterministic methods for 3D blade shape optimization of a 10 kW single-stage radial inflow turbine (RIT) and compares the obtained results with those received from the stochastic or deterministic methods. Eight algorithms were used altogether, including one stochastic, three deterministic and four hybrid algorithms. Principal component analysis (PCA) was used to analyze the optimization process. 3D models of selected reference and optimized geometries were created to compare the differences in the obtained flow characteristics. At least two different geometries were found for which the efficiency increased by above 2 pp. (validated on refined grids). The increased efficiency was obtained over the entire investigated range of mass flow rate, with a value of the total-to-static efficiency of 90.6% at the nominal point obtained using a hybrid method.