Analysis of Thermoelectric Composites for Power Generation
MetadataShow full item record
Thermoelectric composites are promising for energy conversion between heat and electricity. However, due to the complexity of nonlinear thermoelectric coupling, the effective behavior of the thermoelectric composites have not been explored much. The goal of this dissertation is to understand the effective behavior of the thermoelectric composites and to explore the possibility of improving their performance, excluding the size and interface effects. For this purpose, we studied three types of thermoelectric composites: one-dimensional and two-dimensional periodic thermoelectric composites with macroscopic homogeneity, and one-dimensional functionally graded materials. In the first part of this dissertation, we adopt the non-linear asymptotic homogenization theory developed by Yang1 and redefine the effective properties to understand the macroscopic behavior of one-dimensional thermoelectric composites. We discover that the effective behavior of the composites can be described by a single-phase material with homogenized properties, though the effective electric conductivity becomes temperature-dependent. We also find that although the effective figure of merit cannot be higher than both of its constituents, the effective power factor can be substantially improved. In the second part of the thesis, we seek an improved the conversion efficiency by matching compatibility factor with reduced current density in a functionally graded material. The results show that the enhancement by the functionally graded strategy is insignificant because as we match the compatibility with reduced current density, the effective figure of merit is inevitably sacrificed. In the last part of the thesis, we try to understand the behavior of two-dimensional thermoelectric composites. We first examine the asymptotic homogenization theory developed by Yang and note its limitations, and then we develop a direct finite element method as an alternative solution. Through the analysis, we found that the effective behavior of the two-dimensional thermoelectric can also be described by a homogenized single-phase material, and its effective properties do correlate with its conversion efficiency. Lastly, we develop a machine learning based methodology to speed up the search for better thermoelectric composites and reduced more than 95% of the computational cost from FEA. Through this study, we found no improved effective figure of merit from the combinations of the 16 state-of-the-art p-type thermoelectric materials at a wide temperature range. Nonetheless, the developed finite element method and machine learning methodology are proved to be effective in estimating the effective behavior of two-dimensional thermoelectric composites, which can greatly accelerate the new composite discovery and design.
- Mechanical engineering