Accurate prediction of the Higher Heating Value (HHV) of biomass is critical for optimizing bioenergy conversion processes. This study presents a comprehensive data-driven framework that integrates six advanced machine learning models—Multilayer Perceptron (MLP), Radial Basis Function (RBF), Support Vector Regression (SVR), Gaussian Process Regression (GPR), Multiple Linear Regression (MLR), and Adaptive Neuro-Fuzzy Inference System (ANFIS)—to predict HHV using elemental composition (carbon, hydrogen, oxygen, nitrogen, sulfur) and biomass type. Model performance was rigorously evaluated using RMSE, MAPE, EF, and a composite performance score. Among all models, the RBF network achieved the highest accuracy with an overall RMSE of 0.52, MAPE of 2.05 %, and EF of 0.98, outperforming GPR, MLP, and ANFIS. Sensitivity analysis revealed that carbon content is the most influential variable, with its exclusion increasing RMSE from 0.55 to 1.20, confirming its dominant role in HHV prediction. Furthermore, a Genetic Algorithm (GA) was employed to optimize biomass compositions, resulting in significant theoretical enhancements in HHV: for example, industry wastes increased from 24.08 to 44.49 MJ/kg, and briquettes/charcoals reached up to 43.33 MJ/kg. Response surface analysis demonstrated that HHV increases with higher carbon and hydrogen content while decreasing with oxygen. These findings highlight the superiority of RBF and GPR models in HHV prediction and demonstrate the potential of AI-driven optimization to guide the design of high-energy biomass fuels. The study provides a robust, quantifiable modeling approach to support efficient biomass selection and upgrading in waste-to-energy systems.

Keywords