Machine learning-based high-throughput screening, strategical design and knowledge extraction of Pt/Ce_xZr_1−xO₂ catalysts for water gas shift reaction

Hydrogen production through water-gas shift reaction (WGSR) is a paramount option for a sustainable hydrogen economy. ZrO₂-incorporated CeO₂ has been widely applied to WGSR using Pt-based catalysts because of its high oxygen storage capacity, reducibility, and CO oxidation. However, there are challenges in identifying the optimal reaction conditions and structure of Pt/Ce_xZr_1−xO₂ catalysts due to a lack of understanding of the complex catalysis in WGSR over Pt/Ce_xZr_1−xO₂. In this study, we developed a machine learning-based approach to predict the performance of Pt/Ce_xZr_1−xO₂ catalysts using experimental data collected from the literature. Artificial neural networks (ANNs) were exploited to predict the CO conversion of the catalysts in a wide range of WGSR conditions. The ANN model analyzed 79 860 cases, including 110 catalysts, and provided the optimal catalyst structures for maximizing CO conversion in different operating conditions such as a specific reaction temperature, given steam-to-CO ratios, and limited Pt loading level. In addition, the extended analysis of the CO oxidation capability of Ce_xZr_1−xO₂ as support reveals major governing properties that determine the performance of Pt/Ce_xZr_1−xO₂ catalysts in WGSR, from support reducibility at a low-temperature to thermal stability at a high-temperature.