Fully-coupled thermal-electric modeling of thermoelectric generators

Abstract

Numerical models of thermoelectric generators require quantification of discretization uncertainty, the scrutinization of thermoelectric phenomena on model energy imbalances and simultaneous thermal–electric predictions, and the rectification of disagreement with analytic models when considering temperature-dependent material properties. Within this methods paper, two fully coupled, thermal–electric unicouple-level models are developed and evaluated over various thermal and electrical conditions to address the aforementioned issues—a numeric model in ANSYS CFX and an iterative analytic model. Model results were compared to ANSYS Thermal–Electric (TE) and ANSYS Fluent. Agreement between all models’ electrical and thermal predictions was achieved, albeit ANSYS Fluent’s thermal predictions exhibited high percent differences (7–8%) and had global energy imbalances on the order of 15% due to incongruent thermal–electrical power output predictions. ANSYS TE congruently predicts power output when considering the device’s thermal behavior and electrical performance separately, with disagreement on the order of a percent. Through incorporating all thermoelectric phenomena, ANSYS CFX’s global energy imbalances were hundredths to thousandths of a percent; exclusion of pertinent thermoelectric phenomena such as Thomson and Bridgman heating caused imbalances of tens of percent. The inclusion of Thomson heat is imperative when modeling thermoelectric devices. The analytic model’s thermal and electrical performance predictions are within ANSYS CFX’s uncertainty (2–5%), and these predictions yielded percent difference of less than 1% in comparison to ANSYS CFX when the unicouple produces ±50% maximum power. By using temperature-integrated averages of material properties, analytic modeling is sufficient for the thermal–electric characterization of unicouples with interconnectors operating under Dirichlet thermal boundary conditions.