Influence of Biomass Composition and Microwave Pyrolysis Conditions on Biochar Yield and its Properties: a Machine Learning Approach

The investigation of microwave pyrolysis behavior and interactive effects of process parameters through machine learning is necessary to systematically determine the combined effects on the yield and characteristics of biochar. This study involves the prediction of microwave biochar yield and its property using various machine learning approaches. Based on the input data of feedstock characteristics (elemental and proximate composition) and operating conditions of microwave pyrolysis (microwave power, time, weight, absorber), the output targets like biochar yield and higher heating value (HHV) have been predicted. The results suggested that eXtreme Gradient Boosting (XGB) model with optimal hyper-parameters could predict the yield and HHV of microwave-derived biochar with higher correlation coefficient (R-2) of 0.91. The impact of each factor on output target and their interactions during microwave pyrolysis has been observed from SHAP (SHapley Additive exPlanations) dependence plots. The study outcome revealed that microwave power is the most significant feature influencing the yield of biochar and its property (HHV). The present work gives an insight through computational approach in improving microwave pyrolysis of biomass for enhanced biochar yield and its properties.

Automated summary

This study utilizes machine learning approaches to predict the yield and characteristics of microwave-derived biochar, revealing that microwave power is the most significant factor influencing biochar yield and properties.