Design of Selective TPV Thermal Emitters Based on Bayesian Optimization Nesting Simulated Annealing

It is vital to further improve the design of TPV thermal emitters since the energy efficiency of thermophotovoltaic (TPV) systems is still not adequately high. In this paper, we propose a novel evaluator for the optimization of TPV thermal emitters, namely the percentage of effective figure (PEF) to replace the figure of merit (FOM). The associated algorithm, Bayesian optimization nesting simulated annealing (BOnSA), is developed to achieve better performance. By searching throughout the whole parameter space and then optimizing in a reduced space, BOnSA can lead to a satisfactory solution numerically for GaSb photovoltaic (PV) cells. When designing the emitter, the aperiodic material structure with an anti-reflection substructure and Fabry–Perot etalon is constructed from the material candidates. In particular, one of the optimal structures determined by BOnSA is {SiO 2 , ZnS, Ge, MgF 2 , W, Si, SiO 2 , W} with the value of PEF = 0.822 , which is better than the previous work by comparison. Moreover, by applying BOnSA to various structures, we have obtained higher values of PEF with less time cost, which thus verifies the efficiency and scalability of BOnSA. The results of our paper show that BOnSA provides an effective approach to the thickness optimization problem and that BOnSA is applicable in other relevant scenarios.