Evaluating machine learning models for predicting the storage performance of polypyrrole-based supercapacitor electrodes

A.E. Abd-elnaby, Khaled E. El-Kelany, F. Selim, Mostafa A. Ebied, A.M. Elshaer,
Evaluating machine learning models for predicting the storage performance of polypyrrole-based supercapacitor electrodes,
Journal of Energy Storage,
Volume 140, Part A,
2025,
118943,
ISSN 2352-152X,
https://doi.org/10.1016/j.est.2025.118943.
(https://www.sciencedirect.com/science/article/pii/S2352152X25036564)
Abstract: Supercapacitors have emerged as vital energy storage devices, playing a crucial role in advancing sustainable energy solutions by enabling efficient energy storage and rapid charge-discharge cycles. This study investigates the application of machine learning for predicting the electrochemical behavior of polypyrrole-based supercapacitors fabricated using electro-polymerization techniques. The supercapacitors, denoted as Ppy:KNO₃/GRs@100, were prepared by electrodepositing polypyrrole onto graphite sheets over 100 cycles. The fabricated devices were characterized using scanning electron microscopy to analyze surface morphology, X-ray Diffraction for structural analysis, Electrochemical Impedance Spectroscopy to study ion transport and charge transfer properties, and cyclic voltammetry to evaluate electrochemical performance. The dataset used for machine learning was collected from cyclic voltammetry curves recorded at scan rates of 2, 5, 10, 20, and 50 mV/s. Five machine learning algorithms, including random forest, K-Nearest neighbors, ridge regression, lasso regression, and gradient boosting regression, were trained and evaluated using performance metrics such as mean absolute error (MAE), root mean square error (RMSE), and R-squared (ROlabi et al. (2022)2). Gradient boosting regression (GBR) consistently outperformed the other models, demonstrating superior predictive accuracy across all scan rates. For instance, at 50 mV/s, gradient boosting regression achieved an RMSE of 3.2999, MAE of 0.6175, and R2 of 0.96, significantly surpassing other models such as RF and KNN. GBR also maintained R2 values above 0.85 across all scan rates, even as electrochemical complexity increased at higher speeds. The integration of advanced machine learning techniques, particularly gradient boosting regression, offers significant potential for accurately modeling and predicting supercapacitor behavior under diverse conditions. By combining precise fabrication, comprehensive characterization, and robust predictive modeling, this research provides valuable insights into the optimization of supercapacitor design and operation, paving the way for improved energy storage technologies.
Keywords: Artificial intelligence; Supercapacitor; Current prediction; Machine learning; Gradient boosting regression algorithm