Modeling of Dry Reforming of Methane Using Artificial Neural Networks

The process of dry reforming methane (DRM) is seen as a viable approach for producing hydrogen and lowering the atmospheric concentration of carbon dioxide. Recent times have witnessed notable advancements in the development of catalysts that enable this pathway. Numerous experiments have been conducted to investigate the use of nickel-based catalysts in the dry reforming of methane. All these reported experiments showed that variations in the catalyst property, namely pore size, pore volume, and surface area, affect the hydrogen production in DRM. None of the previous studies has modeled the surface nickel-incorporated catalyst activity based on its properties. In this research, DRM's hydrogen yield is predicted using three different artificial neural network-learning algorithms as a function of the physical properties of Ni-based catalyst along with two reaction inputs. The geometric properties as an input set are a different approach to developing such empirical models. The best-fitting models are the artificial neural network model using the Levenberg-Marquardt algorithm and ten hidden neurons, which gave a coefficient of determination of 0.9931 and an MSE of 7.51, and the artificial neural network model using the scaled conjugate gradient algorithm and eight hidden layer neurons, which had a coefficient of determination of 0.9951 and an MSE of 4.29. This study offers useful knowledge on how to improve the DRM processes.