Research on prediction of energy density and power density of biomass carbon-based supercapacitors based on machine learning

Abstract

The advancement of computer technology has made machine learning models widely used to study the electrochemical performance of supercapacitors, thus accelerating material discovery and performance optimization. From the perspective of biomass raw material characteristics, this study innovatively predicts the energy density and power density of biomass carbon-based supercapacitors based on Extreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), and Deep-Learning Neural Network (DNN) models. The results show that the LightGBM model performs best in energy density prediction, with R² reaching 0.922. The XGBoost model has the best effect on the power density prediction, and the R² is as high as 0.984. At the same time, through the analysis of SHAP value, it is found that biomass raw materials' composition and activation conditions are important characteristics affecting energy density and power density, which is very important for optimizing the performance of carbon materials. Therefore, it is a feasible method to predict supercapacitors' energy and power density from the perspective of the characteristics of biomass raw materials. This provides a reliable and valuable method for optimizing the performance of supercapacitors and predicting other performance parameters.