Characterization of selected biomass type and analysis of torrefaction process for production a valorised fuel

Abstract

Lignocellulosic biomasses are promising renewable energy source, but its properties limit direct application. Torrefaction improves lignocellulosic biomass quality and enables direct application for valorisation. Two-step kinetic model was developed to explain biomass thermal degradation under torrefaction process conditions. This study provides methodology for estimating intrinsic kinetics (individual temperature rate constants and global kinetic parameters) using particle swarm optimization (PSO) algorithm from TGA isothermal torrefaction data. It also bridges gab between apparent kinetics observed during fixed bed furnace torrefaction and intrinsic kinetics using scaling factors (estimating using PSO) to account for heat and mass transfer resistances for both feedstocks, wood chips (WC) and palm kernel shell (PKS). In TGA, isothermal torrefaction was performed at 275, 300, and 325 ºC after each for 100 min and in furnace, isothermal torrefaction was performed at same temperature for 15, 30, 45, and 60 min to capture time-resolved solid yield dynamics for both feedstocks. PKS kinetic parameters and solid yields indicate higher thermal stability and structural recalcitrance as it requires more activation energy for primary decomposition compared to highly WC. Fuel properties, such as higher heating value (HHV), oxygen-tocarbon atomic ratio (O/C), and hydrogen-to-carbon atomic ratio (H/C), were effectively linked (R² > 0.98) using weighted least regression (WLS) linear regression with torrefaction severity index (TSI), representing combined effect of process conditions and mass loss. Using apparent kinetic model, TSI – fuel properties WLS model and uncertainty propagation, a robust physics informed synthetic dataset was generated. This dataset was used to train Gaussian process regression (GPR), support vector regression (SVR) and artificial neural network (ANN). All three models showed higher predictive strength against experimental data and predicted optimum process conditions using multi-objective optimization. Overall, this study provides a comprehensive framework to unite robust kinetic modelling with uncertainty-aware machine learning to provide scalable tool for torrefaction process optimization.