Sierra J.D. , Hernández-Salazar C.A. , González-Estrada O.A.
No
J. Phys. Conf. Ser.
Conference Paper
Científica
0
0
01/01/2025
2-s2.0-105027370885
Perovskite solar cells have recently attracted significant attention due to their favorable photoelectric properties. In this study, five different machine learning models, Long Short-Term Memory, Random Forests, Support Vector Machines, Gradient Boosting and eXtreme Gradient Boosting, were trained in Python using a set of categorical input data related to the fabrication of perovskite cells. This data included information on electron and hole transport layer materials, the photovoltaic perovskite used and its deposition process, electronic back contact materials, anti-solvent, and precursor solution. The goal was to perform regression tasks with cell efficiency as the target variable. After evaluating the different models, the random forest model was identified as the most suitable among those studied due to its ability to fit the experimental data and its efficient runtime. While this model offered a reasonable characterization of perovskite solar cells based on energy conversion efficiency and manufacturing parameters, it was limited in its ability to identify optimal fabrication conditions to maximize efficiency. Further studies using alternative approaches are recommended to develop models capable of performing this task more effectively. © Published under licence by IOP Publishing Ltd.
Adaptive boosting; Conversion efficiency; Learning systems; Nanostructured materials; Perovskite; Random forests; Solar cell efficiency; Support vector machines; Electron transport layers; Electrons and holes; Gradient boosting; Machine learning models; Perovskite coatings; Photoelectric property; Random forests; Short term memory; Solar cell performance; Support vectors machine; Perovskite solar cells