Machine learning assisted bio-oil deoxygenation under methane environment: The model compound study on guaiacol

Abstract

The novel approach of methane-assisted deoxygenation on bio-oil is highly promising based on our previous studies. In this study, guaiacol was chosen as the model compound to further screen the catalyst to achieve optimal performances aiming to realize hydrogen-free deoxygenation while enhancing the yield of valuable BTX (benzene, toluene, and xylene). A catalyst of multi-metals doped ZSM-5 was predicted using machine learning model and fabricated to meet the requirements. Additionally, the synergistic effect between the support material and the loading metals were taking into account during the rational design and fabrication process of the catalyst. ZSM-5 was chosen as the support material to achieve a proper acidity to trigger the conversion of guaiacol while maintaining excellent selectivity towards the target products: benzene, toluene, and xylene (BTX). Zn and Ga was co-coped to enhance the methane activation at low temperatures, while Ce was further introduced to prohibit the over coking. A batch mode reactor was applied in the reaction performance studies aiming to understand the conversion, coke yield, selectivity of desired products, etc. The optimized Zn-Ga-Ce/ZSM-5 catalyst showed a conversion of guaiacol with the liquid yield of 58 wt% and BTX selectivity of as high as 90 %. Comprehensive characterizations including NH₃-TPD, DRIFTS, XRD, TEM, BET, and TGA were conducted to gain in-depth understanding on the physicochemical properties of the catalyst. The relationship between the reaction performance and physicochemical properties of the catalyst has been explored carefully. This study will shed light on the catalyst engineering for the methane-assisted conversions.