Optimal coupling design for organic Rankine cycle and radial turbine rotor using CFD modeling, machine learning and genetic algorithm

Abstract

The organic Rankine cycle (ORC) as a potential technology can be used to recover low-grade waste heat. This study provides a rapid and systematic methodology to design ORC system coupling with the radial turbine rotor. Firstly, the coupled model of ORC and the radial turbine is established, and the thermodynamic and economic performance is obtained to optimize the preliminary design parameters. Then, the CFD analysis for the turbine rotor is performed to generate the databases, and the machine learning algorithms (random forest, XGBoost, decision tree, and support vector machine) are used to train the surrogate models. And finally, the optimal shape of the meridional surface is obtained to improve ORC’ power output. The results indicated that the system net power output and the cost rate at the decision point achieved 393.06 kW and 13.59 $/h, respectively. The support vector machine exhibited the highest value of R2 (coefficient of determination) among the regression algorithms. The optimized results showed a more uniform pressure distribution is founding on the suction surface of the optimized structure compared with the basic structure’s, leading to the power output was increased by 7.5 %. The findings can provide references for the rapid coupling design and optimization of ORC and the turbine.