Journal
ENVIRONMENTAL SCIENCE & TECHNOLOGY
Volume 51, Issue 18, Pages 10881-10892Publisher
AMER CHEMICAL SOC
DOI: 10.1021/acs.est.7b01413
Keywords
-
Categories
Funding
- U.S. National Science Foundation [CBET 0955124]
Ask authors/readers for more resources
The complex interactions that occur in mixed species bioelectrochemical reactors, like microbial fuel cells (MFCs), make accurate predictions of performance outcomes under untested conditions difficult. While direct correlations between any individual waste stream characteristic or microbial community structure and reactor performance have not been able to be directly established, the increase in sequencing data and readily available computational power enables the development of alternate approaches. In the current study, 33 MFCs were evaluated under a range of conditions including eight separate substrates and three different wastewaters. Artificial Neural Networks (ANNs) were used to establish mathematical relationships between wastewater/solution characteristics, biofilm communities, and reactor performance. ANN models that incorporated biotic interactions predicted reactor performance outcomes more accurately than those that did not. The average percent error of power density predictions was 16.01 +/- 4.35%, while the average percent error of Coulombic efficiency and COD removal rate predictions were 1.77 +/- 0.57% and 4.07 +/- 1.06%, respectively. Predictions of power density improved to within 5.76 +/- 3.16% percent error through classifying taxonomic data at the family versus class level. Results suggest that the microbial communities and performance of bioelectrochemical systems can be accurately predicted using data-mining, machine-learning techniques.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available