Optimizing Hydrogen Production in the Co-Gasification Process: Comparison of Explainable Regression Models Using...

Area of Science:

• Chemical Engineering
• Sustainable Energy Technologies
• Machine Learning Applications

Background:

• Co-gasification of biomass and plastic waste is a key strategy for producing hydrogen-rich syngas to meet growing clean energy demands.
• Optimizing co-gasification for maximum hydrogen yield is challenging due to diverse feedstocks and process complexities.
• Existing machine learning (ML) models for co-gasification lack consensus on effectiveness and interpretability, especially with limited data.

Purpose of the Study:

• To model the co-gasification process using seven distinct ML algorithms.
• To develop a framework for evaluating ML model interpretability in co-gasification.
• To identify the most suitable ML model for optimizing hydrogen production from co-gasification.

Main Methods:

• Conducted comprehensive experiments assessing generalization ability, predictive accuracy, and interpretability of seven ML algorithms.
• Employed Support Vector Regression (SVR) and compared its performance against other models.
• Integrated Shapley Additive Explanations (SHAP) for detailed feature importance analysis and model interpretability.

Main Results:

• Support Vector Regression (SVR) demonstrated superior performance with the highest coefficient of determination (R2) of 0.86.
• SVR effectively captured non-linear dependencies and mitigated overfitting compared to other models.
• SHAP analysis provided unprecedented insights into feature importance, confirming ML model feasibility for industrial hydrogen production.

Conclusions:

• SVR is identified as the most effective ML model for optimizing biomass and plastic co-gasification for hydrogen production.
• The study establishes a robust framework for evaluating ML model interpretability and performance in this domain.
• Findings support the advancement of sustainable energy technologies and reduction of greenhouse gas emissions through optimized co-gasification.