Interpretable machine learning guided experimental optimization of Perovskite Solar Cells using standardized laboratory data
Won-Jung Jung a, Dong-Geon Kwun b, Hye-Seon Park c, Min-Su Cho c, Ji-Youn Seo b
a Nano-Bio Convergence Research Institute, Pusan National University, Busan, Republic of Korea
b Department of Nano Fusion Technology, Pusan National University, Busan, Republic of Korea
c Department of Semiconductor Materials, Devices and Processes, Pusan National University, Busan, Republic of Korea
Poster, Won-Jung Jung, 110
Publication date: 22nd October 2026

The performance of metal halide perovskite solar cells (PSCs) is strongly influenced by various experimental parameters, including material compositions, device architectures, and fabrication processes. Although substantial experimental efforts have been devoted to improving device efficiency and stability, the systematic organization and utilization of accumulated experimental data remain challenging, particularly in individual research laboratories. Herein, we establish a structured experimental dataset comprising over 1,000 PSC devices fabricated in a single research laboratory, incorporating detailed information on material compositions, device structures, fabrication conditions, and photovoltaic characteristics. To ensure data consistency and reliability, the collected experimental records are systematically cleaned, standardized, and transformed into machine-readable features for machine learning (ML) analysis. Various ML algorithms are employed to predict key photovoltaic performance parameters, including power conversion efficiency (PCE), open-circuit voltage (Voc), short-circuit current density (Jsc), and fill factor (FF), from material properties and fabrication conditions. Furthermore, Shapley additive explanations (SHAP) are utilized to identify the experimental parameters that contribute most strongly to the predicted device performance and to investigate their influence on photovoltaic characteristics. Based on these findings, targeted experiments are designed to evaluate the effects of the identified parameters and experimentally validate the ML-derived insights. This study aims to demonstrate an integrated approach combining systematic experimental data collection, ML-based performance prediction, interpretable feature analysis, and experimental validation, providing a foundation for data-driven optimization of high-performance PSCs.

© FUNDACIO DE LA COMUNITAT VALENCIANA SCITO
We use our own and third party cookies for analysing and measuring usage of our website to improve our services. If you continue browsing, we consider accepting its use. You can check our Cookies Policy in which you will also find how to configure your web browser for the use of cookies. More info