Search

Perovskite solar cells have emerged as a promising frontier in the realm of renewable energy due to their notable attributes of high efficiency and cost-effectiveness. This study delves into the intricate domain of vapor-deposited perovskite solar cells with the primary objective of optimizing their performance through advanced predictive modeling. State-of-the-art machine learning techniques, including random forest, least absolute shrinkage and selection operator, and deep learning, were strategically employed to comprehensively explore critical parameters such as Voc, Jsc, fill factor, perovskite layer thickness, thin-film deposition methods, temperature, and power conversion efficiency. The research findings underscore the exceptional accuracy of deep learning models, revealing a coefficient of determination of 0.91 and minimal mean absolute errors of 0.91. This robustly emphasizes the precision and reliability of deep learning in predicting power conversion efficiency compared to conventional numerical methods. The study further identifies thin-film deposition methods and short-circuit current density as pivotal factors influencing power conversion efficiency, offering invaluable insights for performance enhancement. The advanced deep learning model successfully forecasted an impressive peak power conversion efficiency of 25.4%. This noteworthy achievement is complemented by a recommended device architecture meticulously achieved through high vacuum deposition techniques, specifically incorporating ITO/MoO3/TaTm/CsPbI2Br/C60/bathocuproine (BCP)/Ag. [ABSTRACT FROM AUTHOR]

This study presents a comprehensive investigation into water jet injection dynamics in supersonic crossflows, employing a hybrid approach that integrates machine learning techniques, specifically random forest, with traditional discrete phase methods. The aim is to accurately model water jet structures and understand the underlying dynamics for optimizing flow control strategies. Through extensive numerical analysis, the study examines the influence of key parameters such as pressure levels, Weber number, droplet diameter and water jet velocity on penetration depth. Additionally, machine learning techniques are employed to analyze the impact of momentum, mass flow, pressure, Mach number and injection angle on penetration height. The findings reveal intricate interactions between pressure levels and penetration depth, driven by factors such as momentum transfer, evaporation efficiency and shock wave behavior. A direct correlation is observed between Weber number and penetration depth, emphasizing the role of inertial forces in determining penetration characteristics. Droplet diameter and water jet velocity emerge as critical factors affecting penetration depth, with smaller droplets and higher velocities resulting in deeper penetration into the crossflow. Machine learning analysis highlights the significance of momentum in influencing penetration height, while also indicating comparable effects of pressure, Mach number and injection angle. The random forest model demonstrates robust performance, achieving an accuracy exceeding 86.7% with an average absolute error of 0.00282, underscoring its reliability in predicting infiltration height. [ABSTRACT FROM AUTHOR]