Water Leakage Detection and Localization using Hydraulic Modeling and Classi cation Approach

No Thumbnail Available

Date

2020-03-10

Journal Title

Journal ISSN

Volume Title

Publisher

Addis Ababa University

Abstract

The distribution of treated water over a water distribution system faces di erent losses. In these distribution systems, a signi cant percentage of water is lost due to leakages. There are plenty of researches that were proposed to tackle this water resource wastage. Most of them, however, focus on large bursts and single leakage detection and localization. This thesis proposed a leak detection and localization approach that uses hydraulic modeling and classi cation approaches for detection, and a statistical approach for localization. The general term used in identifying and locating leaks is Leakage Detection and Localization. The detection system aims at detecting small leakages (i.e., below 5% of the total supply). The localization phase of this thesis also aims at localizing multiple leakages. The proposed approach is applied on Hanoi water network benchmark by doing di erent experiments. A realistic dataset is produced based on a benchmark dataset procedure (i.e., LeakDB). The detection and localization results under well-known calibrated hydraulic conditions are almost perfect (i.e., 100% for detection and 90 for localization). The introduction of demand and noise uncertainty, however, a ects the accuracy of the detection of small leakages. The localization methodology has shown e ectiveness in locating two and three leaks that are presented simultaneously in 9 of 20 cases on average accurately, and 11 of 20 cases at least one of them.

Description

Keywords

Leakage Detection, Localization, Hydraulic Modeling, Classification, Combined residuals

Citation