Novel composite could vastly improve thermophotovoltaic systems

Thermophotovoltaic systems (TPV), which can convert solar radiation, fuels, and industrial waste heat into energy, are emerging as an important power generation technology. At the core of these systems are selective emitters that enable high energy conversion efficiency and output power density.

To improve the design of selective emitters and achieve higher performance, Yu et al. developed a neural network that optimizes material and structural parameters simultaneously. The network relies on a deep reinforcement learning-based optimization method that only needs a limited database of input materials. Testing a library of four high temperature resistant materials, the authors found an optimal selective emitter.

“We obtained the optimal selective emitter composed of a novel material combination of TiO₂, Si, and W substrate, which has never been reported before,” said author Run Hu. “This makes us most excited.”

The composite showed significant improvement over previous materials, reaching an energy conversion efficiency of over 38 percent. The authors hope the results can be a first step to creating improved selective emitters for TPV systems.

“However, further study is needed, because long-term high temperature stability remains the primary obstacle hindering the practical utilization of TPV systems,” Hu said. “In addition, we hope that the deep reinforcement learning method we used can bring inspiration to researchers in other fields, including applied physics, nanophotonics, and metamaterial.”

The author plan to continue their work by fabricating a selective emitter from the novel composite and test it in a TPV system.

Source: “Enhancing overall performance of thermophotovoltaics via deep reinforcement learning-based optimization,” by Shilv Yu, Zihe Chen, Wentao Liao, Cheng Yuan, Bofeng Shang, and Run Hu, Journal of Applied Physics (2024). The article can be accessed at https://doi.org/10.1063/5.0213211 .