SCAPS-1D Embracing Machine Learning Modelling of Plasmonic Perovskite Solar Cells: Innovative Structure of Active Sheet and Electron Transport Substances

Perovskites possess exceptional optical and electrical properties, making them promising candidates for significantly enhancing solar cell efficiency. In this simulation study, we utilized the Solar Cell Capacitance Simulator to numerically investigate the performance of solar cells based on the FTO/ETL/AL/Spiro-OMeTAD/Au configuration. The analysis focused on various perovskite active layers—FAPbI_₃, CsGeI_₃, and CsGeI_₂Br—and electron transport layers—SnO_₂, TiO_₂, and WO_₃. Key parameters studied included layer thickness, doping density, and defect density. The simulation results revealed that the optimal power conversion efficiency of 21.22% was achieved using WO₃ as the electron transport layer, with a thickness of 0.23 μm, a defect density of 1 × 10¹⁵ cm⁻³, and a doping density of 5 × 10²⁰ cm⁻³. Additionally, for the perovskite active layer, a CsGeI_₃ composition with a thickness of 0.65 μm, a defect density of 5 × 10¹⁴ cm⁻³, doping density of 1 × 10¹⁶, and a bandgap energy of 1.363 eV demonstrated superior performance, delivering a power conversion effecincy of 28.1%. This exceeded the performance of FAPbI₃ (bandgap energy, 1.51 eV) and CsGeI_₂Br (bandgap energy, 1.579 eV). These findings suggest that the FTO/WO_₃/CsGeI_₂Br/Spiro-OMeTAD/Au structure, particularly with optimized WO₃ and CsGeI₃ layers, holds great potential for high-efficiency solar cell fabrication. Furthermore, machine learning models with random forest algorithim predicted the performance metrics of the investigated solar cells with an accuracy of 88.6%, and power conversion effeciency prediction with R² of 0.6234 underscoring the potential of machine learning in optimizing solar cell design and performance.