Utilization of machine learning algorithms in estimation of syngas fractions and exergy values for gasification of biomass-lignite mixtures in fixed and fluidized bed gasifiers

• Aspen Plus sensitivity analysis examined biomass-lignite blend relationships. • Key gasification parameters, such as ER and temperature, were thoroughly analysed. • A dataset of 6,750 samples was created using 9 biomass-lignite blends in Aspen Plus. • Four ML methods DT, SVR, GPR, and RF were applied to predict syngas and exergy. • A robust model shows strong performance even with limited dataset sizes. Earth’s environmental challenges, such as climate change and pollution, require urgent emission reductions. A thermochemical method that transforms carbon-rich substances into syngas, biomass gasification produces clean hydrogen as a sustainable energy carrier. This process ensures high carbon conversion efficiency while minimizing greenhouse gas emissions. This study examines the gasification of nine biomass-lignite blends using fluidized-bed and fixed-bed gasifiers. A wide range of biomass samples blended with lignite enabled the analysis of different sample characteristics and their impact on the gasification technique. ASPEN Plus® simulations assess the effects of biomass-to-lignite ratio, equivalence ratio (ER), steam to biomass ratio (SBR), and reactor temperature on syngas fraction and system efficiency. Machine learning models gaussian process regression (GPR), random forest (RF), support vector machine (SVM), and decision tree (DT) predict syngas and product gas exergy values, providing a data-driven optimization approach. For hazelnut shell validation, R2 values were 0.98 for the fixed-bed model and 0.96 for the fluidized-bed model. The Random Forest algorithm demonstrated the highest accuracy (R2 = 0.93), outperforming other models. The study also analysed the amount of data required and demonstrated robust models capable of learning with limited data. Since a significant portion of the machine learning process involves dataset creation, the ability to learn from small datasets is crucial. This highlights the significance of data-efficient learning in machine learning applications. Findings contribute to advancing biomass gasification for cleaner hydrogen production.