Global sensitivity analysis and optimization of a multistage thermoelectric generator based on failure probability

Thermoelectric generator (TEG) is a thermoelectric conversion device that mainly utilizes the Seebeck effect, and the electrical energy generated is usually directly converted from thermal energy. However, research on the global sensitivity analysis and optimization of three-stage TEG systems is currently lacking. In this study, a multistage semiconductor TEG system model based on functional failure probability and a global sensitivity analysis model were established. These models are first used to compare the performance of TEG configurations with different stages. Then, based on the double random uncertainties of design variables, a series of studies were conducted. The results showed that the performance parameters of the all TEG configurations exhibited similar performance variation trends. An first-best working current was identified to obtain the first-best output power and conversion efficiency. Next, as the sample size used to determine the variable uncertainty for a three-stage TEG increased, it increased to 4 x 106, the sensitivity indicators of the output response (P and eta) tended to constant values of 0.02322 and 0.02039. Among the 11 considered design parameters, the uncertainties of the heat and cold source temperatures had the most significant influence on the functional failure probability of the three-stage TEG system, followed by the working current, Zeebek coefficient, number of temperature difference components in each stage, and heat transfer coefficient of the cold source. The remaining variables had negligible influence. Finally, small changes in the mean heat and cold source temperatures changed the sensitivity of the output response of the three-stage TEG to each critical design variable, but did not affect the relative ranking of the sensitivity indicators corresponding to each. The results of this study can provide some basis for the further development of TEG systems.