Eventos Anais de eventos
ENCIT 2022
19th Brazilian Congress of Thermal Sciences and Engineering
ESTIMATING THE HIGHER HEATING VALUE OF CANDIOTA COAL USING DIFFERENT METHODS OF MACHINE LEARNING
Submission Author:
Paulo Morgado , MG
Co-Authors:
Paulo Morgado, Bruno Pasa, Nina Salau, Rodolfo Rodrigues
Presenter: Paulo Morgado
doi://10.26678/ABCM.ENCIT2022.CIT22-0617
Abstract
Analyzing coal samples through proximate analysis is a cheaper and less time-consuming method of describing a coal sample composition than the ultimate analysis. It is of interest to try to establish different relationships between coal composition and the higher heating value. Establishing these kinds of relationships, though useful, can sometimes pose some challenges that will require more sophisticated and innovative methods of regression. This work aimed to develop and compare three different methods of machine learning algorithms to develop relationships between the proximate analysis of Brazilian high-ash coal samples and their higher heating value. All three models performed satisfactorily within the limits of acceptable errors, taking into account the heterogeneity of data used in their training. Even in comparison with linear regression, all models performed very similarly, but, only the artificial neural network method was capable of having better metrics than the linear regression.
Keywords
Candiota coal, proximate analysis, machine learning, Higher Heating Value
