Data-Driven\nDevelopment of Heterogeneous Catalysts\nfor Propane Dehydrogenation with Machine Learning and Metaheuristic\nOptimization

Recent advances in data-driven approaches using the machine\nlearning\n(ML) method have enabled the discovery of high-performance materials.\nThis paper presents a hybrid framework that combines ML models with\na metaheuristic optimization algorithm, to explore improved heterogeneous\ncatalysts for propane dehydrogenation (PDH). The framework proposes\nmultiple PDH catalysts, utilizing our laboratory-scale database. A\nunique five-component catalyst, 2.4Ga 2.2Pt 1.7B 1.3Zr/Al<sub>2</sub>O<sub>3</sub>, exhibits superior performance, achieving a\npropylene yield of 58% at 600 °C. This work highlights the excellent\npredictive capability of the framework and offers a new data-driven\napproach for developing high-performance materials for heterogeneous\ncatalysis.