HT Model: Using the Molecular Transformer for Predicting Hydrotreating Reactions

Sudha Cheranma Devi Eswaran, Mariefel V. Olarte, Robert Ralló, Laura N Marrlett, Juliette Harper, Mariella S Anderson, Eric M Shapiro, Patrick M Kozyra, Alexzabria A StarksEnergy 2023 · DOI

Abstract

Sustainable liquid fuel hydrocarbons possess higher energy densities compared to other renewable technologies and thus play a role in the decarbonization of the transportation sector, particularly in long-range applications such as aviation, marine, and rail transport. Through thermochemical pathways to produce liquid hydrocarbons from biomass, catalytic hydrotreatment (HT) is employed to remove heteroatoms, especially oxygen, nitrogen, and sulfur, in the presence of hydrogen. This study used and modified the Molecular Transformer (MT) to investigate its applicability in the product prediction of HT reactions with biomass-derived reactants as starting input. Three different approaches to model implementation were investigated: training the model from scratch using all reactions, fine-tuning a pretrained MT model on a limited HT reaction data set, and modifying the MT model to incorporate reaction conditions as new features. The models used both the simplified molecule-input line-entry system and self-referencing embedded string representations. A combination of a manually created data set consisting of HT reactions mined from peer-reviewed literature and a generic chemical reactions data set from the U.S. Patent and Trademark Office (USPTO) were used to assess the various MT models’ feasibility for HT reactions. The results demonstrated that the models could learn and utilize relevant chemical features, catalysts, and reaction conditions to predict the outcome of the HT reactions. The cumulative accuracy of predicting elementary reactions from a test set consisting of HT-relevant compounds using the modified MT models reached more than 70%, a notable improvement from the initial 20% accuracy with the vanilla MT model. This study represents a significant step toward utilizing deep learning to predict HT reactions and provides valuable insights for future advancements in this field.
BibTeX
@inproceedings{eswaran2023ht,
  author = {Sudha Cheranma Devi Eswaran and Mariefel V. Olarte and Robert Ralló and Laura N Marrlett and Juliette Harper and Mariella S Anderson and Eric M Shapiro and Patrick M Kozyra and Alexzabria A Starks},
  title = {{HT Model: Using the Molecular Transformer for Predicting Hydrotreating Reactions}},
  booktitle = {Energy},
  year = {2023},
  doi = {10.1021/acs.energyfuels.3c02224},
}

Loading related content...

Figure