Using Anomaly Detection Models for monitoring parameters in a Wastewater Treatment Plant

What is it about?

The exponential growth of technology in recent decades has led to the emergence of some challenges inherent to this growth. One of these challenges is the enormous amount of data collected by the different sensors in our society, namely in management processes such as Wastewater Treatment Plants (WWTPs). These infrastructures comprise several processes to treat wastewater and discharge clean water in water courses. Therefore, the concentration of pollutants must be below the allowable emissions limits. In this work, anomaly detection models were conceived, tuned and evaluated to monitor essential parameters such as nitrate and ammonia concentrations and pH to improve WWTP management. Four Machine Learning models were considered, particularly Local Outlier Fraction, Isolation Forest, One-Class Support Vector Machines and Long Short-Term Memory-Autoencoders (LSTM-AE), to detect anomalies in the three parameters mentioned. Through the different experiments, it was possible to verify that, in terms of F1-Score, the best candidate model for the three analyzed parameters was LSTM-AE-based, with a value consistently higher than 97%.

Featured Image

Photo by Chris Ried on Unsplash

Photo by Chris Ried on Unsplash

Why is it important?

In this study, the objective is to conceive, tune and evaluate several candidate models using 4 ML models to detect anomalies in the wastewater management process in a WWTP. Namely, the study focuses on three different parameters in wastewater: nitrate and ammonia concentrations and pH. The ML models designed to detect anomalies in these parameters were based on Isolation Forest (iForest), Local Outlier Fraction (LOF), One-Class Support Vector Machines (OCSVM) and Long Short-Term Memory-Autoencoders (LSTM-AE). The choice of these models for detecting anomalies in the three parameters mentioned also aims to understand and compare the performance of three traditional models widely used in the literature in anomaly detection with a model that takes time series into account. To our knowledge, the application of this model in the context of WWTPs is still a little-explored topic nowadays. Furthermore, another objective was to understand the behaviour of LSTM-AE-based models in terms of performance compared to more traditional ML models for anomaly detection.

Perspectives