Hydrogen Enhancement in Syngas Through Biomass Steam Gasification: Assessment with Machine Learning Models

Artificial intelligence (AI), particularly supervised machine learning, has revolutionized the biofuel industry by enhancing feedstock selection, predicting fluid compositions, optimizing operations, and streamlining decision-making. These algorithms outperform traditional models by accurately handling complex, high-dimensional data more efficiently and cost-effectively. This study assesses the effectiveness of various machine learning algorithms in engineering, focusing on a comparative analysis of artificial neural networks (ANNs), support vector machines (SVMs), tree-based models, and regularized regression models. The results show that random forest (RF) models excel in predicting syngas composition and its lower heating value (LHV), achieving high precision with training and testing RMSE values below 0.2 and R-squared values close to 1. A detailed SHAP analysis identified the steam-to-biomass ratio (SBR) as the most critical factor in these predictions while also noting the significant impact of temperature conditions. This underscores the importance of thermal parameters in gasification and supports the systematic integration of AI in biofuel production to enhance predictive accuracy.