Design and Analysis of Metal Oxides for CO₂ Reduction Using Machine Learning, Transfer Learning, and Bayesian Optimization

Abstract

We aim to achieve resource recycling by capturing and using CO 2 generated in a chemical production and disposal process. We focused on CO 2 conversion to CO by the reverse water gas shift–chemical looping (RWGS-CL) reaction. This reaction proceeds in two steps (H 2 + MO x ⇆ H 2 O + MO x –1 ; CO 2 + MO x –1 ⇆ CO + MO x ) via a metal oxide that acts as an oxygen carrier. High CO 2 conversion can be achieved owing to a low H 2 O concentration in the second step, which causes an unwanted back reaction (H 2 + CO 2 ⇆ CO + H 2 O). However, the RWGS-CL process is difficult to control because of repeated thermochemical redox cycling, and the CO 2 and H 2 conversion extents vary depending on the metal oxide composition and experimental conditions. In this study, we developed metal oxides and simultaneously optimized experimental conditions to satisfy target CO 2 and H 2 conversion extents by using machine learning and Bayesian optimization. We used transfer learning to improve the prediction accuracy of the mathematical models by incorporating a data set and knowledge of oxygen vacancy formation energy. Furthermore, we analyzed the RWGS-CL reaction based on the prediction accuracy of each variable and the feature importance of the random forest regression model.

Integration of Materials and Process Informatics: Metal Oxide and Process Design for CO2 Reduction

“…Experimental data and descriptors of metal oxides, such as Pymatgen 22 and Materials project descriptors, 23 were obtained from a previous report, 15 and regression models were constructed to predict CO 2 and H 2 conversion rates as previously reported. 15 In the case of the CO 2 conversion rate, x represented the experimental conditions and Pymatgen and materials project descriptors. For the H 2 conversion rate, represented the experimental conditions and Pymatgen descriptors.…”

“…Manufacturing the desired product via the RWGS-CL reaction process requires metal oxide and process designs and testing in pilot and actual plants. In recent years, both metal oxide 15 and process designs 16 have been conducted using machine learning. A statistical model y = f ( x ) is constructed using a dataset between x and y , which represent the synthesis conditions and the properties and activities, respectively, and the dataset is an experimental dataset in metal oxide design.…”

A Route Map of Machine Learning Approaches in Heterogeneous CO2 Reduction Reaction

“…This method has been employed to improve the accuracy of the prediction of optimized experimental conditions for metal oxides to achieve target CO 2 and H 2 conversion extents using the knowledge of oxygen vacancy formation energy from a pretrained model. 42 However, the problems encountered by these methods are quite different from the typical supervised learning methods. The utilization of these methods requires a strong correlation between the knowledge of the pretrained source model and the target task domain.…”

“…Later, this active learning method was adopted by others to predict the important descriptors of CO 2 RR for different Cu-based alloy systems. , Transfer learning is another interesting method which utilizes the knowledge from a pretrained model and reuse as a foundation point for another task, reducing the computational cost of producing labeled data. This method has been employed to improve the accuracy of the prediction of optimized experimental conditions for metal oxides to achieve target CO 2 and H 2 conversion extents using the knowledge of oxygen vacancy formation energy from a pretrained model . However, the problems encountered by these methods are quite different from the typical supervised learning methods.…”

Multiobjective Bayesian Optimization Framework for the Synthesis of Methanol from Syngas Using Interpretable Gaussian Process Models

“…The multiobjective optimization for maximizing methanol production and reducing carbon emission uses the genetic algorithm . Using machine learning and Bayesian optimization, they created metal oxides while jointly optimizing experimental parameters to meet the target CO 2 and H 2 conversion predictions . However, to the best of the authors’ knowledge, hardly any work has been reported in the literature that employs interpretable GPR models in a multiobjective Bayesian optimization framework, specifically for syngas-to-methanol conversion.…”