IoT and machine learning approach for the determination of optimal freshwater replenishment rate in aquaponics system
Abstract
Conventional aquaponics conserve water used in aquaponics whereas the literature suggests a certain level of freshwater replenishment or freshwater exchange for good water quality, fish and plant wellbeing, and the overall productivity of the system. This paper deals with the determination of an optimal freshwater replenishment rate for a standard aquaponics system. IoT devices and sensors were used for this project data collection. This paper used linear regressions and ensemble methods to determine the optimal rate of periodic water replenishment to maintain the water quality parameters that determine the yield and productivity of aquaponics systems. Cubic spline and Lagrange interpolation were applied to raw and simulated data. Results were evaluated and compared using statistical error estimation approaches. The best model amongst the investigated machine learning models was gradient boost with an optimal replenishment rate of 19L per week and a water quality of 4.86 for an aquaponic tank of 100 L capacity. The error estimations were a Mean Squared Error of 0.0224, Mean Absolute Error of 0.1137, Root Mean Squared Error of 0.1499, and R2 of 0.7208. This was within 1% of the value obtained from raw and interpolated data using a polynomial regression. These findings suggest that the water quality of an aquaponics system can be maintained at the desired optimal level with a weekly 19% water replenishment, thereby contributing to the improvement of productivity and resource efficiency.Citation
Chandramenon P, Gascoyne A and Tchuenbou-Magaia F ( 2024) IoT and machine learning approach for the determination of optimal freshwater replenishment rate in aquaponics system. Frontiers in Sustainable Resource Management, 3: 1363914.Publisher
Frontiers MediaJournal
Frontiers in Sustainable Resource ManagementAdditional Links
https://doi.org/10.3389/fsrma.2024.1363914Type
Journal articleLanguage
enDescription
© 2024 The Authors. Published by Frontiers Media. This is an open access article available under a Creative Commons licence. The published version can be accessed at the following link on the publisher’s website: https://doi.org/10.3389/fsrma.2024.1363914ISSN
2813-3005EISSN
2813-3005Sponsors
This project has been partially internally funded by the School of Engineering, Computing, and Mathematical Sciences at the University of Wolverhampton and supported by the EU Horizon 2020 MSCA RISE Project ReACTIVE Too, Grant Agreement No 871163.ae974a485f413a2113503eed53cd6c53
10.3389/fsrma.2024.1363914
Scopus Count
Collections
Except where otherwise noted, this item's license is described as https://creativecommons.org/licenses/by/4.0/