Machine learning prediction of biochar yield based on biomass characteristics

Slow pyrolysis is a widely used thermochemical pathway that can convert organic waste into biochar. We employed six machine learning models to predictively model 13 selected variables using pearson feature se-lection. Additionally, partial dependence analysis is used to reveal the deep relationship between feature variables. Both the gradient boosting decision tree and the Levenberg-Marquardt backpropagation neural network achieved training set R-2 > 0.9 and testing set R-2 > 0.8. But the other models displayed lower performance on the testing set, with R-2 < 0.8. The partial dependence plot demonstrates that pyrolysis conditions have greater impact on biochar yield than biomass composition. Furthermore, the highest treatment temperature, being the sole consistently changing feature, can serve as a guiding factor for regulating biochar yield. This study high-lights the immense potential of machine learning in experimental prediction, providing a scientific reference for reducing time and economic costs in pyrolysis experiments and process development.

Automated summary

This study employed machine learning models to predict and model 13 selected variables and revealed the relationships between variables through partial dependence analysis. The gradient boosting decision tree and Levenberg-Marquardt backpropagation neural network performed well in predicting biochar yield. The highest treatment temperature can serve as a guiding factor for regulating biochar yield.