Machine learning-guided optimization of lignocellulose-derived biochar precursors with tailored properties for energy storage applications

Sustainable electrode materials are increasingly needed to support renewable energy storage. In this study, we present an AI-based framework for designing biochar with specialized properties suitable for battery anodes through optimized biomass pyrolysis. Using 474 data samples, the framework focuses on lignocellulosic composition and processing conditions instead of expensive and redundant ultimate analysis, improving its practical applicability. An eXtreme Gradient Boosting Regression (XGBR) model has been created to achieve high accuracy in predicting biochar yield, surface area, and pH values (R = 0.80–0.98; d = 0.88–0.99). SHapley Additive exPlanations values were evaluated to show temperature, biomass sample’s lignin content, and ash content to be significant in determining biochar properties. Multi-objective optimization with the NSGA-II algorithm confirmed safe and feasible biochar processing conditions (temperatures of 620–650 °C) according to energy storage requirements. Validation tests for furfural residues showed high accuracy in biochar preparation with actual surface area and pH of 644 m 2 g -1 and 9.3, respectively. A software application for applying cutting-edge AI-based design strategies for scaling high-quality renewable energy materials was also developed for practical implementation.