Predicting the bandgap and efficiency of perovskite solar cells using machine learning methods

Rapid and accurate prediction of bandgaps and efficiency of perovskite solar cells is a crucial challenge for various solar cell applications. Existing theoretical and experimental methods often accurately measure these parameters; however, these methods are costly and time‐consuming. Machine learni...

Full description

Saved in:
Bibliographic Details
Published in:Molecular informatics 2024-02, Vol.43 (2), p.e202300217-n/a
Main Authors: Khan, Asad, Kandel, Jeevan, Tayara, Hilal, Chong, Kil To
Format: Article
Language:English
Subjects:
AdaBoostRegressor
Band Gap
CatBoostRegressor
Computational efficiency
Efficiency
Energy gap
Experimental methods
GradientBoostingRegressor
KneighborsRegressor
Learning algorithms
Machine Learning
Perovskite Solar Cells
Perovskites
Photovoltaic cells
Predictions
Solar cells
SVR
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Rapid and accurate prediction of bandgaps and efficiency of perovskite solar cells is a crucial challenge for various solar cell applications. Existing theoretical and experimental methods often accurately measure these parameters; however, these methods are costly and time‐consuming. Machine learning‐based approaches offer a promising and computationally efficient method to address this problem. In this study, we trained different machine learning(ML) models using previously reported experimental data. Among the different ML models, the CatBoostRegressor performed better for both bandgap and efficiency approximations. We evaluated the proposed model using k‐fold cross‐validation and investigated the relative importance of input features using Shapley Additive Explanations (SHAP). SHAP interprets valuable insights into feature contributions of the prediction of the proposed model. Furthermore, we validated the performance of the proposed model using an independent dataset, demonstrating its robustness and generalizability beyond the training data. Our findings show that machine learning‐based approaches, with the aid of SHAP, can provide a promising and computationally efficient method for the accurate and rapid prediction of perovskite solar cell properties. The proposed model is expected to facilitate the discovery of new perovskite materials and is freely available at GitHub (https://github.com/AsadKhanJBNU/perovskite_bandgap_and_efficiency.git) for the perovskite community.
ISSN:1868-1743
1868-1751
DOI:10.1002/minf.202300217