| 초록 |
CO2 direct hydrogenation for methanol production is one of core technologies for sustainable CO2 utilization for value-added chemical production. Especially, the discovery of catalysts for promoting a high CO2 conversion and methanol selectivity is essential to facilitate CO2 direct hydrogenation technology. However, it is hard to fully understand complex catalysis system through experimental trial-and-error approach. In this study, we develop a machine learning (ML) practice to understand the fundamental principle of catalyst in the CO2 direct hydrogenation. Based on data from literature, we build various ML models and apply them to predict the catalysts performances (e.g., conversion and selectivity), K-fold validation method was implemented for model evaluation to avoid overfitting. Finally, with the prediction model, we discussed the relationship between catalyst features and catalyst performance. Our preliminary results indicate that the ML-based catalyst performance prediction model can provide understanding and knowledge for catalyst design, thereby leading to useful strategies to a researcher who works on the catalysis and catalyst experiment at an early R&D stage. |
