Precision biochar yield forecasting employing random forest and XGBoost with Taylor diagram visualization

Area of Science:

• * Biomass conversion and renewable energy technologies.
• * Application of machine learning in chemical engineering and materials science.

Background:

• * Pyrolysis is a key waste-to-energy conversion process generating valuable products like biochar.
• * Biochar yield is influenced by feedstock properties and pyrolysis conditions, necessitating accurate prediction methods.
• * Traditional experimental methods for yield prediction are time-consuming and resource-intensive.

Purpose of the Study:

• * To develop and compare data-driven predictive models for biochar yield using machine learning.
• * To identify the most influential feedstock parameters affecting biochar yield.
• * To overcome the limitations of empirical modeling through advanced computational approaches.

Main Methods:

• * Employed five machine learning algorithms: Lasso regression, Tweedie regression, random forest, XGBoost, and Gradient boosting regression.
• * Utilized historical experimental data for training and testing predictive models.
• * Applied SHAP (SHapley Additive exPlanations) based on cooperative game theory to interpret model predictions and identify feature importance.

Main Results:

• * XGBoost outperformed other models, achieving high predictive accuracy with R-squared values of 0.9739 (training) and 0.8875 (test).
• * The XGBoost model exhibited low errors: 2.14% (training) and 3.8% (test) mean absolute percentage error.
• * SHAP analysis revealed that feedstock properties such as ash and moisture content negatively impact biochar yield, while fixed carbon (FPT), nitrogen, and carbon content positively influence it.

Conclusions:

• * Data-driven machine learning models, particularly XGBoost, offer a powerful and accurate alternative to empirical methods for predicting biochar yield.
• * Understanding the influence of feedstock characteristics is crucial for optimizing biomass conversion processes and maximizing biochar production.
• * Precision prognostic technologies enhance biomass logistics, conversion efficiency, and the overall utilization of biomass as a renewable energy source.