Prediction of specific capacitance of activated carbon electrode for energy storage device by machine learning based approach

The best possible storability of a supercapacitor can be achieved by the properly selecting of the electrode materials. The optimized result can be obtained by applying a Machine Learning (ML) algorithm to predict the best possible electrode material and its storability. In this study, ML algorithms are employed to predict the specific capacitance of activated carbon electrode-based supercapacitors, with a focusing on optimizing their performance. A comprehensive dataset of supercapacitors with their physiochemical features is obtained from previously published research articles. It is used for training and validation of various ML models for prediction of Specific capacitance of supercapacitor. The various physiochemical features which have a bigger impact on storability, are analysed in this study including specific surface area (SSA), pore size, pore volume, heteroatom doping of material, potential window and Id/Ig ratio. By integrating various ML algorithms, including regression and classification methods, the study identified the most significant parameters influencing supercapacitor performance. Different ML models like random forest, decision tree, linear regression, and XG Boost were applied in this study. The result demonstrates that the random forest model, with a root mean square error (RMSE) of 61.88 and a coefficient of correlation (R²) value of 0.84 shows better results, as compared to other exhibits prediction in comparison to other ML models. The feature analysis, cross-correlation analysis in the dataset indicate that the SSA, nitrogen doping and pore volume can considerably influence the performance of supercapacitors. The results when compared to other literature show better results which depict that ML can act as a powerful tool for accelerating material design and optimizing electrode characteristics for future energy storage devices. The predictive models developed in this research could minimize the time needed for the design and assessment of electrode materials for supercapacitors.