Machine learning-driven optimization of Ni-based catalysts for catalytic steam reforming of biomass tar

Abstract

Biomass gasification is a promising process for producing syngas, which is widely used in various industrial processes. However, the presence of tar in syngas poses a significant challenge to biomass gasification due to the difficulties in its removal and potential downstream issues, such as clogging, slagging, and corrosion. Extensive efforts have been made to address this challenge through catalytic tar removal using various catalysts, generating a vast amount of experimental data. Processing this large dataset and gaining new insights into process optimization requires the development of efficient data analysis methods. In this study, a comprehensive database was built, encompassing a total of 584 data points and 14 input parameters collected from literature published between 2005 and 2020. Machine learning algorithms were then trained using this dataset to predict and optimize the catalytic steam reforming of biomass tar. The predicted results were found to agree well with the experimental data. The results show that the reaction temperature is the most important process parameter, with the highest relative importance of 0.24, followed by the support (0.16), additive (0.12), nickel (Ni) loading (0.08), and calcination temperature (0.07), among the 14 input parameters. This work has proposed optimal ranges for the reaction temperature (600–700 °C), Ni loading (5–15 wt%), and calcination temperature (500–650 °C). Furthermore, it was found that a larger specific surface area and higher Ni dispersion are two critical factors for selecting additives and supports. This study provides insights into key parameters for optimizing the catalytic steam reforming of biomass tar, enabling enhanced efficiency and effectiveness in biomass gasification processes.