机器学习筛选乙酰丙酸加氢制-戊内酯催化剂

打开文本图片集

中图分类号：TQ426.6；TP181 文献标识码：A

Abstract： As a key derivative of levulinic acid，a biomass platform molecule， Y-valerolactone holds significant application value in the synthesis of sustainable fuels and chemicals. However， traditional catalyst screening methods are characterized by low eficiency and high costs. To address this，an algorithmic framework was developed for the rapid identification of catalysts that yield high conversion rates and product yields. Initially， the K-Means clustering algorithm classified the dataset， followed by the application of SMOTE and ADASYN oversampling techniques to balance the data. Four machine learning models，Adaboost，Random Forest，Support Vector Machine，and Neural Network were then trained. The two top-performing models，Support Vector Machine and Neural Network were selected based on the training outcomes. Subsequently， the SHAP algorithm analyzed feature importance，identifying reaction temperature and Ru/Pt active metals as key influencing factors. Leveraging these insights，a multi-objective optimization genetic algorithm determined the optimal catalyst. The resultant Ru/N@ CNTs catalyst enhanced both levulinic acid conversion and γ -valerolactone yield，demonstrating improved efficiency and reduced costs compared to traditional methods. This established machine learning framework effectively resolves challenges in catalyst screening and optimization， while the integration of machine learning with experimental data elevates the accuracy of catalytic performance predictions and screening efficiency.

Keywords： machine learning; clustering algorithm;oversampling technique; model optimization hydrogenation catalyst

催化剂的性能直接决定化学反应效率和产品质量，因此高效筛选催化剂是化工领域的核心挑战。（剩余8582字）