Leakage detection and localization in water distribution networks with multiple inlets
Translated title
Lækagedetektering og -lokalisering i vand distributionsnetværk med flere vandforsyninger
Authors
Nielsen, Esben Hovkjær ; Christensen, Mark ; Gnap, Simon Nymann
Term
4. term
Education
Publication year
2021
Submitted on
2021-06-03
Abstract
Denne afhandling undersøger, hvordan lækager kan detekteres og lokaliseres i vanddistributionsnet (WDN) med flere indløb for at reducere vandspild og forbedre effektiviteten. Arbejdet udvikler en ny steady-state model, der udvider eksisterende enkelt‑indløbs modeller til multi‑indløb, og genererer et restsignal som forskellen mellem målte og estimerede tryk. Da restsignalet er støjpræget, anvendes Generalized Likelihood Ratio (GLR) til at detektere ændringer i middelværdien som indikator for lækage. Til lokalisering bruges en følsomhedsmatrix til at generere hypotetiske restsignaler for et læk ved hver node; ved at sammenligne deres retning med det målte restsignal peges sandsynlige lækagesteder ud. Metoden verificeres i simuleringer, mens forsøg på en lille fysisk opsætning (Smart Water Infrastructure Laboratory, SWIL) afslører praktiske udfordringer, sandsynligvis forårsaget af utilstrækkelig parameterestimering, hvilket understreger behovet for omhyggelig parameterbestemmelse og øget robusthed over for modelunøjagtigheder. Derudover designes en MIMO output‑tilstandsfeedback‑controller til at opfylde modellens særlige krav og håndtere kommunikationsforsinkelser i laboratoriet.
This thesis investigates how to detect and localize leaks in water distribution networks (WDNs) with multiple inlets to reduce water loss and improve efficiency. It develops a new steady‑state model that extends single‑inlet approaches to multi‑inlet networks and generates a residual from the difference between measured and estimated pressures. Because the residual is noisy, the Generalized Likelihood Ratio (GLR) is used to detect changes in its mean as a leak indicator. For localization, a sensitivity matrix produces hypothetical residuals for a leak at each node; comparing their direction to the measured residual highlights likely leak locations. The method is verified in simulation, while tests on a small physical setup (Smart Water Infrastructure Laboratory, SWIL) reveal practical issues likely due to poor parameter estimation, emphasizing the need for careful parameter identification and improved robustness to model inaccuracies. In addition, a MIMO output state feedback controller is designed to meet model‑specific requirements and handle communication delays in the laboratory.
[This summary has been generated with the help of AI directly from the project (PDF)]
Keywords
Lækage ; Control ; Fejldetektering ; SWIL
Documents