Predicting CO2 adsorption in KOH-activated biochar using advanced machine learning techniques.

Area of Science:

• Materials Science
• Environmental Science
• Chemical Engineering

Background:

• Accurate carbon dioxide (CO2) adsorption forecasting in KOH-activated biochar is vital for geoengineering and environmental technologies.
• Understanding adsorption is key to developing efficient carbon capture solutions.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting CO2 adsorption in KOH-activated biochar.
• To identify key factors influencing CO2 adsorption using various ML techniques.

Main Methods:

• A diverse set of ML algorithms including Support Vector Regression (SVR), CatBoost, Random Forests, and Neural Networks were employed.
• Models were trained and validated on a dataset of 329 data points, with performance assessed using metrics and visualizations.
• Monte Carlo outlier detection and Taylor Diagrams were used for dataset validation and model performance analysis.

Main Results:

• SVR and CatBoost models demonstrated the highest predictive accuracy for CO2 adsorption.
• SVR achieved an R² of 0.9235 and MSE of 0.2207, while CatBoost achieved an R² of 0.9327 and MSE of 0.1942.
• Sensitivity and SHAP analyses identified pressure and temperature as critical parameters influencing CO2 adsorption.

Conclusions:

• Advanced ML models, particularly CatBoost and SVR, are highly effective for predicting CO2 adsorption in biochar.
• These findings provide valuable insights for optimizing industrial carbon capture processes and future research.
• The study highlights the potential of ML in enhancing adsorption efficiency for environmental applications.