Machine learning-aided catalyst screening and multi-objective optimization for the indirect CO2 hydrogenation to methanol and ethylene glycol process†

Abstract

Indirect CO₂ hydrogenation to methanol and ethylene glycol is a green, efficient, and economical technique for converting CO₂ into high-value chemicals to address the intractable environmental crisis caused by CO₂ emissions. However, traditional methods for screening and optimizing catalysts in this process mainly depend on experience and repeated ‘trial-and-error’ experiments, which are resource-, time- and cost-consuming tasks. Therefore, this study developed a machine learning framework for predicting the conversion ratio of ethylene carbonate and the yield of methanol and ethylene glycol from the indirect CO₂ hydrogenation technology to accelerate the catalyst screening and optimization processes. The initial dataset was visualized by conducting principal component analysis and improved to ensure sufficient information variables for the machine learning model to distinguish between catalyst types. After comparing the optimized results of three algorithms, the neural network with two hidden layers is the core of the machine learning model of the indirect CO₂ hydrogenation process. It was then further optimized by a feature engineering method coupled with feature importance analysis and the Pearson correlation matrix. It indicates that the optimized neural network model has higher performance, especially in predicting ethylene carbonate conversion and product yields. Compared with other input features, the space velocity and hydrogen/ester ratio are the two most important factors affecting the conversion ratio of ethylene carbonate and the yield of methanol and ethylene glycol. Based on the results of the feature importance analysis, a multi-objective optimization model with a genetic algorithm was employed to screen the most suitable catalyst. Compared with other catalysts, more efforts should be devoted to the optimized xMoO_x–Cu/SiO₂ catalyst for the industrialization of indirect CO₂ hydrogenation technology after experimental verification.