Enhancing Infrared Solar Absorption Efficiency Through Plasmonic Solar Absorber Using Machine Learning-assisted Design

This research introduces the architecture of an infrared solar energy absorber coupled with absorption prognosis employing machine learning techniques. Our approach involves creating an efficient absorber tailored for infrared wavelengths complemented by a machine learning model for accurately predicting absorption levels. The absorber's design focuses on maximizing absorption within the 0.7 mu m to 4.0 mu m range. We optimized the absorber's parameters, including resonator thickness, substrate thickness, and angle of incidence. Simulation results demonstrate excellent absorption performance, capturing over 90% of light within the specified range. At angles between 0 degrees and 40 degrees, the average absorptance exceeds 80%, peaking at 97.16%. However, at an 80 degrees angle of incidence, absorptance drops to 23.3%. The study employs a 1D-CNN regression model to estimate absorption at various wavelengths, which greatly decreases the time required for simulations and experiments. The findings demonstrate the promise of combining metamaterial structures with machine learning approaches to boost the efficiency of solar energy harvesting and conversion processes.