Investigation of prediction approaches for the design and performance analysis of supercapacitors with biomass-based activated carbon electrodes

Abstract

This article presents an innovative approach integrating biomass valorization and artificial intelligence for developing high-performance supercapacitor electrodes. Apricot kernel shells were converted into activated carbon through controlled hydrothermal carbonization at varying durations (24, 36, and 48 h) followed by KOH activation. 48-hour processed sample exhibited exceptional characteristics with a BET surface-area of 976.650 m2/g and superior electrochemical performance. Comprehensive characterization through SEM, EDS, XRD, and FTIR analyses revealed the evolution of porous structures and surface chemistry modifications during activation process. Electrochemical testing demonstrated that optimized electrode achieved a specific capacitance of 210 F/g at 1 A/g current density, with excellent charge-discharge symmetry and reversibility. A novel aspect of this configuration is the development of an XGBoost-based machine learning framework for predicting supercapacitor performance, achieving an R2 score of 0.991 and RMSE of 0.042. Proposed predictive model, validated through extensive residual analysis and normal probability assessment, enables rapid screening of potential electrode materials. Successful integration of sustainable biomass processing with advanced machine learning techniques establishes an ecosystem for efficient development of next-generation energy storage devices. Findings demonstrate the viability of agricultural waste valorization for high-performance supercapacitor applications while introducing a computational approach that significantly accelerates the optimization of electrode materials.