Predictive modeling and advanced statistical approaches for enhancing biodrying efficiency in wet refuse-derived fuel

• Optimized biodrying reduced final moisture content from 60% to 30%. • GLMM models captured 85% variance in biodrying outcomes. • Aeration rates and initial moisture content were key biodrying factors. • Interaction models reduced prediction error by 20%. • Advanced statistical methods enhanced biodrying process insights. Effective management of wet refuse-derived fuel (RDF) through biodrying is crucial for advancing the circular economy and sustainable waste management. This study aims to optimize the biodrying process of RDF by comparing multiple regression analysis (MRA) and multilevel analysis using generalized linear mixed models (GLMM). Experimental data based on different aeration rates (0.2–0.8 m³/kg feedstock/day) and initial moisture content (MC) (40–60%) were analyzed to predict the final MC. The results indicate that the interaction models significantly outperformed non-interaction models, with GLMM explaining 86% of the variance (R² = 0.86) and reducing the prediction accuracy by 11%. The GLMM framework effectively captured batch-to-batch variability, leading to an optimal final MC reduction from 60% to 30%. Advanced statistical techniques can thus refine waste enhancement processes, providing important insights into biodrying optimization. Improving energy recovery from waste may contribute to establishing more sustainable waste management practices.