Selection of Ni-based catalysts for CO2 methanation using data processing via regression models (machine learning)
machine learning; catalysis; methanation; CO2 hydrogenation; nickel based catalyst
The CO2 methanation process is a crucial step in the energy transition towards a carbon-free future. By utilizing carbon dioxide as a renewable feedstock in catalytic hydrogenation reactions to produce methane, not only does it enable the recycling of carbon resources, but it also meets the growing demand for natural gas. Thus, catalytic hydrogenation of CO2 into methane stands out as an attractive defossilization technology in the fight against climate change, as it consumes CO2 with H2 derived from renewable energy sources to generate CH4. However, predicting the performance of catalytic systems for this process remains a challenge. The application of Machine Learning techniques emerges as a promising approach to predict the performance of potential catalysts based on experimental descriptors. For this purpose, a database of Ni-based catalysts was created from literature data and subjected to data mining. Ensemble tree-based machine learning regression models were developed to predict CO2 conversion and CH4 selectivity based on experimental descriptors. In this context, the significance of variables such as Ni content, calcination temperature, and reaction temperature in predictive modeling was highlighted. The experimental validation of the prediction model and the interpretation of the insights generated not only confirmed the effectiveness of the method but also established this approach as a fundamental tool to guide and optimize future experiments.