Design and analysis of a machine learning-optimized multi-layered absorber for renewable energy applications

Harvesting solar energy into other useful energy sources are essential now a days, hance there are many systems available for that, such as solar absorbers, solar PV, etc. In this study, we have investigated four different solar absorber designs. Out of that four, the Hash-Shaped Solar Absorber (HSSA) structure is more effective for absorbing solar radiation and converting it into heat energy. With the broad response range this structure absorb more than 90% of the radiation in observed range (100-5000 nm). The HSSA structure demonstrates absorption > 99 % at specific wavelength peaks, including 170 nm, 550 nm, 1790 nm, 2750 nm, 3090 nm, and 3700 nm within the solar spectrum. With this the HSSA structure is having more than 93 % absorption in the 3000-4000 nm range, whereas more than 97 % absorption is achieved in the visible region. Further, the HSSA structure is treated with the Machine Learning model as the parameter optimization. The greatest R-2 value for this structure is 0.999592 with a test size of 0.25, and the mean squared error is 1.79801 x 10(-4). The ML reduces time (takes 1/4 of the traditional time) and other simulation requirements. Additionally, the structure is polarization insensitive and up to 60 degrees incident angle insensitive. Due to these characteristics, the HSSA structures find usage in various applications, including solar thermal energy harvesting systems, plasmonic sensors, and detectors.