Bayesian-optimized infrared grating for tailoring thermal emission to boost thermophotovoltaic performance

Abstract

Thermophotovoltaic (TPV) devices, which can break the Shockley–Queisser limit (33.7%) and enhance the thermal energy utilization efficiency, have garnered increasing attention in recent decades. Structuring the emitter surface has been demonstrated to be powerful for tailoring thermal emission to enhance the power density and system efficiency of a TPV system. However, the design and optimization of the broad parameters of the surface nanostructures manually remain to be thorny issues. In this paper, the Bayesian algorithm under the framework of material informatics was coupled with a rigorous coupled wave analysis to optimize the geometry of the infrared grating nanostructure to achieve wavelength-selective emission to boost the TPV performance. It is demonstrated that only less than 0.173% of the total candidate structures were calculated to find out the optimal structure with high spectral emittance in the range of 0.3–1.708 μm, and the power density and system efficiency of the TPV system were enhanced to 4.20 W/cm2 and 35.37%, respectively. The present machine-learning-based optimization of a multi-parameter nanostructure can improve the performance of the TPV system significantly and can be extended to other physical fields in a feasible manner.