Detecting anomalies in water distribution networks using EPR modelling paradigm

Abstract

Sustainable management of water distribution networks (WDNs) requires effective exploitation of available data from pressure and flow devices. Nowadays, water companies are collecting a large amount of such data and they need to be managed correctly and analysed effectively using appropriate techniques. Furthermore, water companies need to balance the data gathering and handling costs with the benefits of extracting information useful for making reliable operational decisions. Among different approaches developed in the last few decades, those implementing data mining techniques for analysing pressure and flow data appear very promising. This is because they can automate mundane tasks involved in the data analysis process and efficiently deal with the vast amount of, often imperfect, sensor data collected. Furthermore, they rely on empirical observations of a WDN behaviour over time, allowing reproducing/predicting possible future behaviour without employing hydraulic simulation of the network, which require continuous/iterative calibration of the model based on on-line fresh data. This paper investigates the effectiveness of the evolutionary polynomial regression (EPR) paradigm to reproduce and predict the behaviour of a WDN using on-line data recorded by low-cost pressure/flow devices. Using data from a real district metered area (DMA), the case study presented in this paper shows that by using the EPR paradigm a model can be built which enables to accurately reproduce and predict the WDN behaviour over time and detect flow anomalies due to possible unreported bursts or unknown increase of water withdrawal. Such an EPR model might be integrated into an early warning system to raise alarms when anomalies are detected.

An Integrated Bottom-Up Approach for Leak Detection in Water Distribution Networks Based on Assessing Parameters of Water Balance Model

“…Datadriven approaches focus on mining the burst-induced features from vast historical data, and then the leakage detection problem can be turned into a classification or clustering problem [19,20]. Laucelli et al investigated the effectiveness of the evolutionary polynomial regression paradigm to reproduce the behavior of a WDN using online data recorded by low-cost pressure/flow devices [21]. Palau et al applied principle component analysis (PCA) to the control of water inflows into DMAs of urban networks [22].…”

A comprehensive comparative analysis of deep learning tools for modeling failures in smart water taps

“…The study revealed that the LSTM model effectively handles the shortcomings of ordinary artificial neural networks when predicting complex conditions in Smart Water Networks. In their work, Muharemi et al (2018) presented some approaches for abnormal event detection on aquatic time series data. Wang et al (2019) found that a deep learning method based on neural networks outperforms the support vector machine (SVM).…”

Disturbance Extraction for Burst Detection in Water Distribution Networks Using Pressure Measurements

“…For example, around 7.85 × 10 9 m 3 of water was lost in 2017 in China (MOHURD, 2017). In addition, frequent pipe bursts may threaten public health (Fox et al, 2016), damage the urban environment (Laucelli et al, 2016), and disrupt the normal water supply service. Efficient and reliable detection of bursts helps water companies to react more quickly, saving scarce water resources, helping safeguard public health, and improve the sustainability of water supplies.…”