An AI-Based Modelling of a Sorption Enhanced Chemical-Looping Methane Reforming Unit

Hydrogen as a green fuel has attracted enormous attention recently. Although hydrogen combustion produces no harmful by-products, hydrogen production can be almost disastrous. Hydrogen production mainly originates from fossil fuels, and more than 80% of hydrogen production is produced using fossil fuel reformation with CO2 formation as a by-product. Light hydrocarbon gases, predominantly methane, are extensively used for hydrogen production. While methane reforming is an economical and efficient process, decarburization of flue gas can be a challenge. Processes involving chemical looping can be used to mitigate these challenges, and they are favorable for simultaneous CO2 capture during hydrogen generation. Intelligent models can help have accurate monitoring of such plants. The aim of this paper is to provide an Artificial Intelligence (AI) based approach to model a Sorption-Enhanced Chemical-Looping Reforming (SECLR) unit. To this end first, a SECLR unit was simulated using ASPEN Plus version 11. Then the simulation results were validated by experimental data, and the SECLR unit went through 31000 different scenarios. The derived data from ASPEN Plus was modeled and simulated with machine learning methods to estimate the CH4 conversion, H2 Purity, and CO2 removal in the SECLR process. Artificial neural networks, ensemble learning, and support vector machine methods were developed to predict the CH4 conversion, H2 Purity, and CO2 removal in a SECLR unit. All three models could provide satisfactory results for predicting CH4 conversion, CO2 removal, and H2 Purity. According to statistical evaluations, Artificial Neural Network (ANN) outperformed Support Vector Machine (SVM) and ensemble learning in producing results with lower error values and higher accuracy with an average 5.23e-5 of error and R2 of 0.9864.