Predictive Modelling of Graphene-Based Supercapacitors for Enhanced Energy Storage Applications: A Machine Learning Approach

Abstract

Despite the growing interest in graphene-based materials for supercapacitors, owing to their high electrical conductivity and huge specific surface area, there are currently no systematic methods for accurately predicting their electrochemical performance. Current research in this area is often hindered by empirical trial–and–error techniques and fragmented datasets, impeding the logical design and optimization of high-performing devices. To address these gaps, the present study leverages ML to forecast the performance of graphene-based supercapacitors, focusing on specific capacitance, power density and energy density. A comprehensive dataset was compiled from existing literature, encompassing physicochemical properties and electrochemical test features. Three ML models, Gradient Boosting Regression (GBR), Random Forest Regression (RFR), and Multiple Linear Regression (MLR) were employed to predict supercapacitor performance. The GBR model achieved the best overall performance with R² values of 0.9, 0.7, and 0.8, and MSE values of 1.03 F²/g², 5.05 (Wh/kg)², and 2.30 (W/kg)² for specific capacitance, energy density, and power density respectively. The results indicate that GBR outperformed other models, achieving the highest determination coefficient (R²) values and the lowest mean squared error (MSE) for energy density, power density, and specific capacitance. RFR showed comparable robustness with slightly higher MSE values, while MLR had the lowest accuracy among the three. Correlation analysis revealed that annealing temperature and current density significantly influence specific capacitance and power density, respectively. This study underscores the potential of ML in optimizing graphene-based supercapacitors, providing valuable insights for the advancement of next-generation energy storage technologies.