Forecasting Catalytic Property‐Performance Correlations for CO₂ Hydrogenation to Methanol via Surrogate Machine Learning Framework

Abstract

toward a greener energy mix together with more sustainable chemical production is halfway but will still require many more years or perhaps decades and massive investment to pervade the market. Additionally, some sectors such as cement industries intrinsically emit CO 2 . [2] Opting for solutions based on carbon capture and storage (CCS) as well as carbon capture and use (CCU) can help to restrain persisting CO 2 emissions. CCS involves storing carbon within geological formations which is an efficient way to curb CO 2 emissions, but this technology is costly and energy intensive. Hence, CCU is a relatively more promising and attractive option. [3] The captured CO 2 can be used as a renewable resource to manufacture chemicals (formic acid, acetic acid, and carbonates) and fuels (methanol, ethanol, etc.). [4] However, the thermodynamic stability of CO 2 molecule (π-conjugated stable structure) puts forward a critical impediment in CO 2 use and thus, its use as chemical feedstock is presently limited to few industrial processes such as urea production and methane synthesis in Powerto-Methane plants. Moreover, the phase transformation in its conversion process, i.e., final products in liquid phase (e.g., methanol and ethanol), while reactants (CO 2 and H 2 ) in gaseous phase, renders the reaction entropically less favorable. All the above limitations led to the advent of CO 2 conversion in presence of high-performance metal-based catalyst for activating CO 2 molecule under milder conditions. [5] CO 2 hydrogenation over metal-based catalysts can effectively produce methanol (CH 3 OH) which is one of the most promising environment-benign fuels. Although, several approaches for CO 2 conversion to methanol such as electrochemical, photochemical, and thermochemical processes have been investigated but among them, only thermochemical route has given high conversion yields for methanol synthesis. [2,6,7] Methanol can be used as a blending component or replacement for gasoline in IC engines for improving the octane rating and reducing the emissions of SOx, NOx, and hydrocarbons. Also, its storage and transportation is comparatively easier than gasoline as it is less flammable and exists in liquid state at room temperature. [8] Moreover, methanol can be employed as chemical feedstockThe hydrogenation of CO 2 to methanol has been studied by several researchers owing to its two significant merits, namely, mitigation of CO 2 emissions and production of renewable fuel. Consequently, numerous experiments have been reported for carbon-neutral methanol synthesis over copper-based catalysts over the past few years. However, it is very expensive and time-consuming to always observe reaction changes with respect to input parameters (operating conditions and catalytic properties) experimentally. Herein, this study develops an ultrafast machine learning (ML) based framework to predict CO 2 conversion and methanol selectivity by comprehensive knowledge extraction (extensive database construction) from existing published literature. Among...

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

“…The optimization results of the different learning algorithm models are evaluated based on the mean square error (MSE), determination coefficient ( R 2 ), and mean absolute error (MAE) under 5-fold cross-validation, and they are described as follows. 29,30 where n is the number of samples; Y i and Ŷ i stand for the real and predicted values of the i th data; and Ȳ denotes the average value of the dataset.…”

“…The optimization results of the different learning algorithm models are evaluated based on the mean square error (MSE), determination coefficient (R 2 ), and mean absolute error (MAE) under 5-fold cross-validation, and they are described as follows. 29,30 MSE ¼ 1 n…”

Interpretable Machine Learning for Accelerating Reverse Design and Optimizing CO2 Methanation Catalysts with High Activity at Low Temperatures

“…After preparing the data set of the CO 2 methanation process, it is randomly partitioned into two segments with an 80:20 ratio, following a commonly employed approach in ML model development. 20,23 The training set comprises 80% of the data points and is utilized for adjusting hyperparameters of the ML algorithm and training the prediction model. Meanwhile, the remaining 20% of the data points are reserved for evaluating the predictive accuracy of the trained ML model.…”

“…However, it is equally crucial to consider other output variables, such as CH 4 selectivity and CH 4 yield, for a comprehensive evaluation of catalytic performance. For example, Tripathi et al 20 believed that comprehensive consideration of CO 2 conversion ratio and methanol yield can provide a more scientific analysis of the relationship between performance properties and reaction parameters of CO 2 to methanol process. Bonke et al 21 highlighted the potential of ML coupled with multimetric optimization can effectively expedite advancements in various catalysis domains by offering unparalleled insights into performance bottlenecks and enhancing comparability.…”