Machine learning prediction of supercapacitor performance of N-doped biochar from biomass wastes based on N-containing groups, element compositions, and pore structures

Abstract

N-doped biochar has great potential for development in the field of supercapacitors. In this study, Random Forest and Extreme Gradient Boosting models are used to predict the specific capacitance of N-doped biochar. The prediction is based on several features, including pore structure parameters, element composition, N-containing group of N-doped biochar, and electrochemical testing characteristics. Shapley additive explanations and partial dependency plots are used to explore the impact of the features on specific capacitance. Results show that both the Random Forest and the Extreme Gradient Boosting models exhibited excellent prediction performance, with R2 of 0.95 and 0.96, respectively. N-6 contributes more to the higher specific capacitance among the three N-containing groups. According to the partial dependency plots, when the specific surface area, pore size, and degree of graphitization are around 2200 m2/g, 4 nm, and 1, respectively, the specific capacitance of N-doped biochar is about 303 F/g at 1 A/g. In addition, a procedure for predicting the specific capacitance of N-doped biochar is developed based on the PySimpleGUI library and the Extreme Gradient Boosting model. This study provides a reference for the preparation of high specific capacitance of N-doped biochar.