Work

Machine Learning in Perovskite Photovoltaics: A Mini‐Review

Year: 2026
Type: article
Abstract: Perovskite solar cells (PSCs) have emerged as a central focus in photovoltaic research due to their high power conversion efficiency (PCE), solution processability, and excellent charge carrier transport properties. However, material screening, process control, and device optimization still rely heavily on time‐consuming trial‐and‐error experiments and computationally intensive first‐principles simulations, both of which severely limit PCE gains. Machine learning (ML), with its ability to extract patterns from high‐dimensional heterogeneous data and enable cross‐scale predictions, is becoming a key technological pathway to overcome this bottleneck. This article reviews ML applications in PSCs. First, it outlines the typical workflow and five mainstream algorithms: kernel methods, random forest, gradient boosting decision trees, artificial neural networks, and graph neural networks. Then, it evaluates representative achievements across dimensions—material design, process optimization, and device engineering—comparing the applicability, prediction accuracy, and generalization capabilities of different algorithms across various tasks. Finally, it identifies core challenges and proposes future directions, including multimodal fusion modeling, physics‐informed embedding, and closed‐loop intelligent experimental platforms, to support data‐driven, rational design throughout the PSC development pipeline.
Source: cMat
Language: English
Open Access status: diamond
OpenAlex ID: W7235032043
FWCI:
Cites: 84
Cited by:
Related to: 10