Accurate prediction of green hydrogen production based on solid oxide electrolysis cell via soft computing algorithms

Area of Science:

• Energy Science and Engineering
• Materials Science
• Chemical Engineering

Background:

• Solid oxide electrolysis cells (SOECs) are key for converting renewable energy to green hydrogen.
• Existing SOEC models lack generalizability across different systems.
• Accurate modeling is crucial for optimizing hydrogen production efficiency.

Purpose of the Study:

• To develop robust, data-driven machine learning models for SOEC hydrogen production.
• To identify the most accurate and reliable modeling techniques for SOEC systems.
• To understand the influence of various parameters on hydrogen output.

Main Methods:

• Utilized a comprehensive suite of machine learning algorithms including ANNs, CNNs, RFs, SVMs, and ensemble methods like XGBoost.
• Trained and validated models on a dataset of 351 SOEC operational points.
• Employed Monte Carlo outlier detection for dataset validation and SHAP values for sensitivity analysis.

Main Results:

• ANNs, CNNs, Gradient Boosting, and XGBoost models demonstrated superior performance with high R-squared values and low error metrics.
• All input parameters were found to significantly impact hydrogen production.
• Electrode conditions (current and cathode) were identified as critical factors influencing SOEC performance.

Conclusions:

• Machine learning, particularly ANNs, CNNs, Gradient Boosting, and XGBoost, offers a powerful approach for predicting SOEC hydrogen production.
• The developed models provide a reliable framework for optimizing SOEC operations.
• Findings offer valuable insights for future research and industrial applications in green hydrogen generation.