关键词: 乙醇催化转化, 正丁醇, 机器学习, 催化剂筛选, 贝叶斯优化

The conversion of bioethanol to n-butanol represents a crucial pathway for the long-term storage and high-value utilization of renewable energy. Traditional trial-and-error approaches to obtain efficient catalysts for n-butanol synthesis from vast synthetic spaces are time-consuming, labor-intensive, and inefficient, while being incapable of covering extensive material combinations. This study introduces machine learning methodologies to explore a data-driven research paradigm for catalyst screening and design aimed at improving n-butanol yield. We collected literature data on heterogeneous catalysis for ethanol-to-n-butanol conversion since 2000, establishing a comprehensive catalyst dataset encompassing catalyst composition, preparation methods, structural characteristics, reaction conditions, and corresponding catalytic performance. Nine regression algorithms were evaluated based on this dataset, with the eXtreme Gradient Boosting (XGBR) model achieving coefficient of determination (R²) values of 0.986 and 0.897 for the training and test sets, respectively, demonstrating superior predictive performance and generalization capability compared to other models. Coupled with Bayesian optimization, five high-performance candidate catalysts were efficiently identified within the n-butanol synthesis catalyst parameter space. This work provides novel insights for the efficient screening and rational design of catalysts for ethanol-to-n-butanol conversion and opens a data-driven avenue for the high-value and efficient utilization of ethanol.

Key words: ethanol catalytic conversion, n-butanol, machine learning, catalyst screening, Bayesian optimization