Ghinea, Liliana Maria (University Dunarea de Jos Galati), Vasiliev, Iulian (Dunarea de Jos University of Galati), Barbu, Marian (Dunarea de Jos University of Galati)

Deep Learning Techniques Employed for Anomaly Detection in Wastewater Treatment Plants

Scheduled for presentation during the Regular Session "Fault diagnosis I" (WeAC), Wednesday, June 11, 2025, 12:10−12:30, Room B

33rd Mediterranean Conference on Control and Automation, June 10-13, 2025, Tangier, Morocco

This information is tentative and subject to change. Compiled on May 9, 2025

Abstract

Accurate anomaly detection in wastewater treatment plants (WWTPs) is essential to secure operational efficiency, environmental compliance and public health safety. This paper investigates two Deep Learning approaches employed for detecting mechanical faults injected in the Dissolved Oxygen (DO) sensor of a WWTP, namely Convolutional Neural Network (C-NN) and Long Short-Term Memory Neural Network (LSTM-NN). Experimental results demonstrate the strengths and limitations of each approach, with both achieving good levels of accuracy, precision, recall and F1-score under varying operational conditions. However, when evaluating the values of the performance metrics obtained for the two Deep Learning approaches, it is proved that the LSTM-NN is more suitable for the task at hand. This study provides insights into the application of Neural Networks for improving fault detection accuracy and enhancing the reliability of wastewater treatment plants.