Artificial Neural Network-Based Mathematical Model of Methanol Steam Reforming on the Anode of Molten Carbonate Fuel Cell Based on Experimental Research

The article describes a mathematical model of methanol steam reforming taking place at the anode of a molten carbonate fuel cell (MCFC). An artificial neural network with an appropriate structure was subjected to a learning process on the data obtained during an experiment on the laboratory stand for testing high-temperature fuel cells located at the Institute of Heat Engineering of the Warsaw University of Technology. The backpropagation of error method was used to train the neural network. The training data included the results of methanol steam reforming in the fuel cell for steam-to-carbon ratios of 2:1, 3:1, and 4:1. The artificial neural network was then asked to generate results for other steam-to-carbon ratios. As a result, the artificial neural network predicted that the highest power density for a molten carbonate fuel cell working on methanol would be obtained with a steam-to-carbon ratio of 2.8:1. The article’s key achievement is the application of artificial intelligence to calculate an unusual steam-to-carbon ratio for the methanol steam reforming process occurring directly at the anode of an MCFC fuel cell. The solution proposed in the article contributed to reducing the number of experimental studies.