Machine learning and thermodynamic modeling for optimizing hydrogen production via algae-biomass co-gasification

The growing demand for sustainable energy has intensified interest in hydrogen (H 2 ) production via biomass co-gasification. This study uniquely integrates thermodynamic modeling with advanced machine learning (ML) techniques to optimize H 2 production and syngas quality from algae-biomass blends—a novel approach that addresses critical gaps in renewable energy production. Three algae species (Chlorella vulgaris, Nannochloropsis oculata, and Fucus serratus) were co-gasified with three biomass types (Fir Pellet (FP), Palm Empty Fruit Bunch (PEFB), and Pellet Pine Wood (PPW)) using a comprehensive Aspen Plus® simulation framework based on Gibbs free energy minimization. Unlike previous research that typically relies on either simulation or experimental methods in isolation, our approach combines Aspen Plus® simulations with six distinct ML algorithms to create a comprehensive predictive framework. Simulations identified optimal conditions, including a gasification temperature (GT) of 800°C and an equivalence ratio (ER) of 0.1–0.5, achieving energy and exergy efficiencies of 87.73% and 84.79%, respectively. The innovative incorporation of algae significantly enhanced H 2 yield and syngas quality, leveraging algae's high reactivity and sustainability benefits. Six ML models—namely Extreme Gradient Boosting (XGB), Random Forest (RF), Support Vector Regression (SVR), K-Nearest Neighbors (KNN), Decision Tree (DT), and Artificial Neural Networks (ANN)—were applied to predict H 2 yield and syngas lower heating value (LHV). XGB demonstrated unprecedented predictive accuracy for H 2 yield, with an R 2 exceeding 0.98 and RMSE of 0.220, while RF excelled in LHV predictions, achieving an R 2 of 0.951 and RMSE of 0.452. This study is the first to identify and quantify the significance of key operational parameters (GT, steam-to-fuel ratio (S/F), ER, and blend in feed (%) (BF (%))) through Shapley additive explanations (SHAP), establishing a novel methodology for optimizing renewable H 2 production that can be adapted across various feedstock combinations, advancing the field of sustainable energy systems.