Evidential neural network and metaheuristic optimization algorithms for sustainable biomass utilization in bioethanol and bio-based chemical production

Abstract

The efficient conversion of lignocellulosic biomass into fermentable sugars is a crucial step in bioethanol production. This study explores the application of advanced machine learning (ML) models, particularly the Evidential Neural Network (ENN), in predicting and reducing sugar yields from Sida cordifolia and Ipomoea repens. The study compares the performance of ENN, Gaussian Process Regression-Bayesian Optimization (GPR-BO), Support Vector Machine-Particle Swarm Optimization (SVM-PSO), and Artificial Neural Network-Particle Swarm Optimization (ANN-PSO) using identical input variables, including acid concentration, reaction time, and temperature. The results demonstrate that ENN outperforms all other models with the lowest error, indicating perfect predictive accuracy. ANN-PSO also exhibited strong performance goodness-of-fit, while GPR-BO showed moderate predictive capability. SVM-PSO, however, had the lowest accuracy, with significant deviations from observed values. The findings suggest that ENN, combined with metaheuristic optimization techniques, provides a highly reliable predictive framework for biomass applications by effectively managing data uncertainty through Dempster-Shafer theory. The study highlights reducing sugar yield from Sida cordifolia (RSY-SC) as a more efficient feedstock compared to reducing sugar yield from Ipomoea repens (RSY-IR), based on key performance metrics. Despite the promising results, computational complexity and the need for large-scale experimental validation remain challenges for ENN implementation. Future research should focus on hybrid AI models, real-time AI-powered biorefinery systems, and integration with lifecycle assessments (LCA-TEA) to further optimize bioethanol production. These advancements could contribute to sustainable bioenergy solutions, reducing reliance on fossil fuels while enhancing efficiency, accuracy, and economic feasibility in lignocellulosic biomass conversion.