Study on the Co-gasification characteristics of biomass and municipal solid waste based on machine learning

Abstract

Co-gasification of biomass and municipal solid waste (MSW) exhibits synergistic effects by improving the quality of syngas while reducing environmental pollution from MSW. In this study, Machine learning (ML) techniques were employed to investigate the co-gasification process of biomass and MSW. A comprehensive dataset was constructed using existing data, including different feedstock types and operating conditions, with 18 input features and 9 output features. Four advanced ML models were utilized to model and analyze the co-gasification process. By leveraging feedstock characteristics and operating parameters, key gasification parameters such as syngas composition, lower heating value (LHV) of syngas, tar yield, and carbon conversion efficiency were predicted. The results showed that all four models exhibited excellent predictive performance, with R2 values greater than 0.9 in both the training and testing stage. Specifically, Histogram-based gradient boosting regression (HGBR) exhibited the lowest root mean square error (RMSE) in predicting CO, while the gradient boosting regressor (GBR) achieved the best performance in H2 prediction with a RMSE of 1.6. The most influential input features for CO concentration were equivalence ratio (ER), oxygen content in biomass and hydrogen content in biomass. The key features affecting H2 concentration were steam/fuel and ER.