Research on specific capacitance prediction of biomass carbon-based supercapacitors based on machine learning

With the development of computer technology, a variety of machine learning models have been used to predict the electrochemical performance of energy storage devices. In this study, the elemental analysis, industrial analysis and structural composition, activation conditions, and current density of biomass are innovatively selected as input conditions from the biomass raw material characteristics perspective and predict specific capacitance based on the light gradient boosting machine (LightGBM) and deep neural network (DNN) algorithm. Meanwhile, the prediction effects of the contrast capacitance under seven different input combinations are compared. The results show that the combination prediction effect of retaining all features is the best, and the industrial analysis and structural composition of biomass have a greater impact on the model than elemental analysis, this conclusion is also verified by SHapley Additive exPlanations (SHAP) value analysis, indicating that they are essential for the model. It is also found that the Light GBM model has more advantages, with R2 of 0.951, mean absolute error (MAE) of 11.090, and Root-mean-square error (RMSE) of 14.756. This study builds a key bridge between the basic composition information of biomass raw materials and the electrochemical performance. Also, it helps to understand the related formation mechanism of biomass-derived carbon materials.