A hybrid model-based method for leak detection in large scale water distribution networks

Abstract

During the past decades, the problem of finding leaks in Water Distribution Networks (WDN) has been controversy. The quicker detection of leaks prevents water loss and helps avoiding their economic and environmental consequences. On the other hand, increasing the speed of leak detection increases the false leak detection that imposes high costs. In this paper, we propose a real-time hybrid method using AI algorithms and hydraulic relations for detecting and locating leaks and identifying the volume of losses material. The proposed method relies on simple and cost-effective flow sensors installed on each junction in the pipeline network. We demonstrate how influential features for leak detection would be generated by using hydraulic equations like Hazen-Williams, Darcy-

Precise Water Leak Detection Using Machine Learning and Real-Time Sensor Data

“…As said before, that study only used a 200-sample dataset to train their model, which can be why they got almost 40% improvement. When referring to the work in [16], we also conclude that RF is the best algorithm for leak detection. Our results are similar across all algorithms, which helps validate our approach and test.…”

“…The authors of [16] used data from smart meters located alongside water distribution pipelines to train a set of ML algorithms in order to detect leaks. The methodology followed was training the algorithms on different pipes for one day, with some pipes not showing any leakage, and testing the trained models on a different day.…”

“…To improve the efficiency of water distribution and to reduce waste related to leaks, a sensor network can be used to execute those analyses with precision and in an autonomous way. Some research and development can be found in both the academic and industrial worlds, ranging from expansive fixed systems that use ultrasonic clamps to analyze real-time water flow, such as [11], to low-cost solutions found in the academic world, such as [12][13][14][15][16].…”

Viability of Pressure-Reducing Valves for Leak Reduction in Water Distribution Systems

“…Leak location is possible through modeling (Adedeji et al 2019;Liu et al 2021), monitoring (Cheng 2021), PRVs in conjunction with SCADA (Güngör et al 2019), and has implications for time and energy savings (Mysorewala 2019). Leaks can also be found using machine-based learning (Cantos et al 2020), using multiple paths algorithm (Hao Png et al 2020), non-destructive techniques (Aslam et al 2018), deep learning convolutional neural networks (Lei et al 2020), monitoring and occupancy data (de Coning and Mouton 2020), multistage optimal valve operations and smart demand metering (Huang et al 2020), a hybrid model-based method (Fereidooni 2021 et al), modeling-based algorithms (Taghlabi et al 2020), computational intelligence (Quiñones-Grueiro et al 2021), a pressure and data-driven classifier approach (Sun et al 2020), acoustics (Stephens et al 2020), IoT (Thenmozhi et al 2021), an integrated bottom-up approach (Yu et al 2021), a CNN with mel frequency cepstral coefficients (Chuang et al 2019), and pressure analysis using a machine learning ensemble (Fuentes and Pedrasa 2019), among others.…”

Flowrate Time Series Processing in Engineering Tools for Water Distribution Networks

“…Many water distribution networks (WDNs) are continuously monitored through installed sensors that measure hydraulic parameters (e.g., pressure, flowrate, users’ consumption) and water quality parameters (e.g., chlorine concentration, pH, temperature) (Kara et al., 2016). This continuous monitoring allows to collect raw data to be used in multiple engineering applications, being flowrate and pressure data the most widely used time series by water utilities in different engineering applications, such as: the calculation of water balances (Meseguer & Quevedo, 2017); the development and calibration of hydraulic models in terms of nodal demands and pipe roughness coefficients (Do et al., 2016; Zhang et al., 2018; Zhou et al., 2018); the application of burst detection and location techniques by inverse analysis (Blocher et al., 2020; Moasheri & Jalili‐Ghazizadeh, 2020; Sophocleous et al., 2019), by using classifier approaches (Capelo et al., 2021; Fereidooni et al., 2021; Hu et al., 2021), or by using transient‐based techniques (Capponi et al., 2017; Covas & Ramos, 2010; Covas et al., 2004; Duan, 2017). Fiorillo et al.…”