Artificial neural network simulating microbial fuel cells with different membrane materials and electrode configurations

Tsompanas, Michail Antisthenis; You, Jiseon; Wallis, Lauren; Greenman, John; Ieropoulos, Ioannis

Abstract

© 2019 Elsevier B.V. Microbial fuel cells (MFCs) are gaining interest due to higher power production achieved by deep analysis of their characteristics and their effect on the overall efficiency. To date, investigations on MFC efficiency, can only be based on laboratory experiments or mathematical modelling. However, there is only a handful of rule-based mathematical modelling due to the difficulties imposed by the high sensitivity of the MFC system to environmental parameters and the highly complex bacterial consortia that dictate its behavior. Thus, an application of an artificial neural network (ANN) is proposed to simulate the polarisation of cylindrical MFCs with different materials as the separation membranes. ANNs are ideal candidates for investigating these systems, as there is no need for explicit knowledge of the detailed rules that govern the system. The ANN developed here is a feed-forward back-propagation network with a topology of 4-10-1 neurons that approximates the voltage of each MFC at a given state. Two different membrane materials with two different electrode configurations were assembled and utilized in laboratory experiments to produce the data set on which the ANN was trained upon. For the whole data set the correlation coefficient (R) between real values and outputs of the network was 0.99662.

• Journal Article Type: Article
• Acceptance Date: Jun 30, 2019
• Online Publication Date: Jul 9, 2019
• Publication Date: Oct 1, 2019
• Deposit Date: Jul 11, 2019
• Publicly Available Date: Jul 10, 2020
• Journal: Journal of Power Sources
• Print ISSN: 0378-7753
• Publisher: Elsevier
• Peer Reviewed: Peer Reviewed
• Volume: 436
• Article Number: 226832
• DOI: https://doi.org/10.1016/j.jpowsour.2019.226832
• Keywords: artificial neural networks, microbial fuel cells,polarisation curves
• Public URL: https://uwe-repository.worktribe.com/output/1491788
• Publisher URL: http://doi.org/10.1016/j.jpowsour.2019.226832
• Additional Information: Additional Information : This is the author’s accepted manuscript. The published version can be found on the publishers website here: https://doi.org/10.1016/j.jpowsour.2019.226832
• Contract Date: Jul 11, 2019