Evaluation of design parameters of a biomass gasifier by using DeepTabTransformer and multiple linear regression models

Abstract

In present study, effects of various parameters such as; fuel type (low rank coal, leaf trash pellet from maniple solid waste; MSW and a 50/50 mixture of leaf trash pellet and low rank coal), catalysts (bentonite and zeolite as natural catalysts, iron oxide (Fe2O3) and potassium carbonate (K2CO3) as synthetic catalysts), equivalence ratio (ER = 0.25, 0.29, and 0.32), temperature, and airflow rate are experimentally investigated using the thermocatalytic gasification process. Results of the experimental study are extracted by means of carbon conversion efficiency (CCE), cold gas efficiency (CGE), lower heating value (LHV) and converted syngas (H2, CO, CO2, CH4). DeepTabTransformer method, a type of deep learning (DL) regression model, is applied to the results to obtain a comprehensive parametric evaluation. Furthermore, the multiple linear regression method (MLR), a type of machine learning (ML) regression model, is used for model validation. As a result, it was shown that gasifying leaf trash pellets and coal together (co-gasification) with natural/organic catalysts can significantly improve syngas quality, conversion efficiency, and reduce tar formation. From the regression analyses, it is yielded that R2 is 0.941 for syngas lower heating value (LHV), 0.941 for CGE, and 0.994 for CCE. Additionally, lower error values are obtained by DeepTabTransformer model with a root mean square error (RMSE) of 0.1076 MJ/kg for LHV and 0.0102 for CGE, demonstrating its effect in modeling nonlinear relationships between parameters. © 2026 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.