Robust design optimization (RDO) of thermoelectric generator system using non-dominated sorting genetic algorithm II (NSGA-II)

Abstract

The thermoelectric generator (TEG) is a promising technology for the exhaust heat recovery of automobiles and TEG optimization has been widely studied. However, previous TEG optimization studies did not consider variations in the TEG net power output caused by uncertain parameters in the TEG system. This paper introduces a robust design optimization (RDO) that maximizes the mean of the performance function while minimizing its variance, leading to an optimum design that is less sensitive to uncertainties in TEG systems. A surrogate model is used to reduce the computational cost and the non-dominated sorting genetic algorithm II (NSGA-II) is used to find a compromise solution. The standard deviation of the TEG net power output of the deterministic optimum design (93.51W) verifies that the uncertainty of TEG systems significantly affects the variation of the TEG net power output, indicating that the uncertainty should be considered in TEG optimization problems. The compromise solution guarantees stable and high TEG net power output compared to the deterministic optimum design stemming from existing TEG optimization studies. The results of a global sensitivity analysis using the Sobol index indicate that the inlet temperature of the hot fluid has the greatest impact on the TEG net power output.