The rapid advancement of machine learning (ML) technology across diverse domains has provided a framework for discovering and rationalising materials and photovoltaic devices. This study introduces a five-step methodology for implementing ML models in fabricating hole transport layer (HTL) free carbon-based PSCs (C-PSC). Our approach leverages various prevalent ML models, and we curated a comprehensive dataset of 700 data points using SCAPS-1D simulation, encompassing variations in the thickness of the electron transport layer (ETL) and perovskite layers, along with bandgap characteristics. Our results indicate that the ANN-based ML model exhibits superior predictive accuracy for C-PSC device parameters, achieving a low root mean square error (RMSE) of 0.028 and a high R-squared value of 0.954. The novelty of this work lies in its systematic use of ML to streamline the optimisation process, reducing the reliance on traditional trial-and-error methods and providing a deeper understanding of the interdependence of key device parameters.

机器学习 (ML) 技术在不同领域的快速发展为发现和合理化材料和光伏器件提供了框架。本研究介绍了在制造无空穴传输层 (HTL) 的碳基 PSC (C-PSC) 时实施机器学习模型的五步方法。我们的方法利用了各种流行的 ML 模型，并使用 SCAPS-1D 模拟整理了包含 700 个数据点的综合数据集，涵盖电子传输层 (ETL) 和钙钛矿层的厚度变化以及带隙特性。我们的结果表明，基于 ANN 的 ML 模型对 C-PSC 器件参数表现出卓越的预测准确性，实现了 0.028 的低均方根误差 (RMSE) 和 0.954 的高 R 平方值。这项工作的新颖之处在于它系统地使用机器学习来简化优化过程，减少对传统试错方法的依赖，并提供对关键设备参数相互依赖关系的更深入理解。