EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Optimization of Wastewater Treatment Through Machine Learning-Enhanced Supervisory Control and Data Acquisition: A Case Study of Granular Sludge Process Stability and Predictive Control

Igor Gulshin, Olga Kuzina

Automation · 27.12.2024 · 6 (1) · 2

DOI: 10.3390/automation6010002 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Automation, online erschienen am 27.12.2024. Verfasst von Igor Gulshin und 1 weiteren Autor:innen. Die Arbeit wurde bisher 19-mal zitiert (Crossref) und nennt 43 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Research Moscow State University of Civil Engineering.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This study presents an automated control system for wastewater treatment, developed using machine learning (ML) models integrated into a Supervisory Control and Data Acquisition (SCADA) framework. The experimental setup focused on a laboratory-scale Aerobic Granular Sludge (AGS) reactor, which utilized synthetic wastewater to model real-world conditions. The machine learning models, specifically N-BEATS and Temporal Fusion Transformers (TFTs), were trained to predict Biological Oxygen Demand (BOD5) values using historical data and real-time influent contaminant concentrations obtained from online sensors. This predictive approach proved essential due to the absence of direct online BOD5 measurements and an inconsistent relationship between BOD5 and Chemical Oxygen Demand (COD), with a correlation of approximately 0.4. Evaluation results showed that the N-BEATS model demonstrated the highest accuracy, achieving a Mean Absolute Error (MAE) of 0.988 and an R2 of 0.901. The integration of the N-BEATS model into the SCADA system enabled precise, real-time adjustments to reactor parameters, including sludge dose and aeration intensity, leading to significant improvements in granulation stability. The system effectively reduced the standard deviation of organic load fluctuations by 2.6 times, from 0.024 to 0.006, thereby stabilizing the granulation process within the AGS reactor. Residual analysis suggested a minor bias, likely due to the limited number of features in the model, indicating potential improvements through additional data inputs. This research demonstrates the value of machine learning-driven predictive control for wastewater treatment, offering a resilient solution for dynamic environments. By facilitating proactive management, this approach supports the scalability of wastewater treatment technologies while enhancing treatment efficiency and operational sustainability.

Förderung

  • National Research Moscow State University of Civil Engineering

Zitieren

APA

Gulshin, I., & Kuzina, O. (2024). Optimization of Wastewater Treatment Through Machine Learning-Enhanced Supervisory Control and Data Acquisition: A Case Study of Granular Sludge Process Stability and Predictive Control. Automation, 6(1), 2. https://doi.org/10.3390/automation6010002

Vancouver

Gulshin I, Kuzina O. Optimization of Wastewater Treatment Through Machine Learning-Enhanced Supervisory Control and Data Acquisition: A Case Study of Granular Sludge Process Stability and Predictive Control. Automation. 2024;6(1):2. doi:10.3390/automation6010002

BibTeX
@article{Gulshin2024,
  author = {Gulshin, Igor and Kuzina, Olga},
  title = {Optimization of Wastewater Treatment Through Machine Learning-Enhanced Supervisory Control and Data Acquisition: A Case Study of Granular Sludge Process Stability and Predictive Control},
  journal = {Automation},
  year = {2024},
  volume = {6},
  number = {1},
  pages = {2},
  doi = {10.3390/automation6010002}
}
Kostenlos starten →