Prediction of Thermionic Energy Conversion Performance and Parametric Effects Using Genetic Algorithms to Fit Physics-Inspired Model Equations to Prototype Test Data

Abstract

Abstract Several recent studies have generated experimental performance data for narrow-gap thermionic energy conversion devices. This investigation explores the use of genetic algorithm methods to fit existing data from literature with physics-inspired model equations. The resulting model equations can be used for performance prediction for system design optimization, or to explore parametric effects on performance. The model equations incorporate Richardson’s law for current density, including both the saturated and Boltzmann regimes, with appropriate relations for power delivered to the external load. The transition regime is characterized using two separate models, each accounting for non-uniformity in emission surfaces and other irregularities in the manufacturing process that impact the output power density. The trained models enable performance prediction of a small-gap thermionic energy conversion device with inputs of only emitter temperature and load resistance. The prototype data considered here tested a thermionic energy conversion device to determine the output current and output power as a function of diode voltage and the emitter temperature. In this study, the prototype test data is used with postulated work function for the emitter and the collector materials and two additional parameters to characterize the transition region. As a result, these postulated functions are substituted into the physics-inspired models, yielding performance models with three adjustable constants. Optimized values of these constants are determined using a genetic algorithm to best fit the experimentally determined performance data for prototype thermionic conversion devices tested in earlier studies. This process yields four insightful results. First, this work demonstrated two pathways to a model that can accurately predict performance trends. The developed models can also be used for design optimization and provide a pathway to a data-driven performance model. The two performance models each provide a method for extracting work function information from performance data. The extracted work function ultimately proves to be independent of the performance model used, as both models produce the same effective work function values for the prototype device. Finally, both of the final three-regime models illuminate the transition region between the saturated and Boltzmann regime.