ScholarWorks@동국대학교

초록

Correlations are commonly used to estimate the higher heating values of different fuels based on either proximate or ultimate analysis. A large number of existing correlations have been developed for different groups of fuels such as coals or biomass. While only a small number of general correlations have been developed which are suitable for all fuels. More recently machine learning methods have also been adapted to more accurately predict these heating values. However, machine learning approaches are typically harder to reproduce compared to correlations. In this study novel new high order polynomials are developed with the aim to bridge this gap and obtain higher accuracy correlations while avoiding the complexity of machine learning. For a large diverse set of fuels it is found that a cubic equation with 23 parameters predicts higher heating values with an R2 of 0.9376. This is higher than the best existing correlation which gives an R2 value of 0.9088 and slightly lower than the best machine learning method which gives an R2 value of 0.9567. In addition to generally outperforming other correlations this is found to be the most accurate correlation for wood, grass, husks, municipal solid waste and organic residue fuel sub-groups. © 2026 Elsevier Ltd.

키워드