Prediction of the higher heating values of biomass using machine learning methods based on proximate and ultimate analysis

The higher heating values (HHV) of biomass are a key component in the analysis and design of any bioenergy system. Many correlations have been published for the estimation of HHV of biomass based on proximate and ultimate analysis. This study aims to predict the HHV of biomass through the application of machine learning algorithms. Multi-linear regression was employed to establish correlations. The sensitivity of the input parameters was investigated, and eight distinct models were created, four for the proximate analysis and four for the ultimate analysis. The extreme gradient boosting algorithm generated the best outcomes for all models. The highest R2 value of 0.9987 was obtained. Models based on ultimate analysis exhibited better performance than those based on proximate analysis. Moreover, the models created using machine learning techniques outperformed those built using statistical approaches.