Iterative Learning Pressure Control and Leakage Localization in Water Network Distribution
Authors
Sirvidas, Danas ; Revelis, Artis ; Jama, Ibrahim Abdiweli
Term
4. term
Education
Publication year
2022
Submitted on
2022-06-02
Pages
91
Abstract
This project investigates whether an Iterative Learning Control (ILC) approach can be implemented in the AAU Smart Water Infrastructure Laboratory and what effect it has on a water distribution system. ILC lets a controller learn from repeated operation and use that experience to improve the next run. This fits water networks, where consumer demand often follows recurring patterns. To enable testing, specific requirements were defined. The water network was modeled with both a static model (steady-state) and a dynamic model (time-varying). After introducing ILC, multiple tests were carried out, first in simulations and then in the laboratory. A leak detection method was also integrated. It uses pressure residuals—differences between measured pressures from the dynamic system and pressure estimates from the static model—which are compared to a threshold to flag leaks. The findings indicate that an ILC-based controller is feasible to implement in the water lab and could be expanded further in future work.
Dette projekt undersøger, om en iterativ læringsbaseret styring (Iterative Learning Control, ILC) kan implementeres i AAU Smart Water Infrastructure Laboratory, og hvilken effekt den har på et vanddistributionssystem. ILC er en metode, hvor styringen lærer af gentagne forløb og bruger erfaringerne til at forbedre næste kørsel. Det er relevant i et vandnet, fordi forbrugernes brug ofte gentager sig over tid, så systemet ser de samme mønstre igen og igen. I arbejdet blev der fastlagt konkrete krav for at kunne teste styringen. Der blev udviklet både en statisk model (for stationære forhold) og en dynamisk model (for tidsafhængig drift) af vanddistributionsnettet. Efter introduktion af ILC blev der gennemført flere tests, først i simuleringer og derefter i laboratoriet. Som en del af løsningen blev lækagedetektion integreret. Den bygger på trykafvigelser (residualer), beregnet som forskellen mellem målte tryk fra den dynamiske model og estimerede værdier fra den statiske model, som sammenlignes med en tærskelværdi for at identificere lækager. Resultaterne peger på, at en ILC-baseret styring kan implementeres i vandlaboratoriet og har potentiale til at blive udbygget yderligere fremover.
[This apstract has been rewritten with the help of AI based on the project's original abstract]
Keywords
ILC ; Iterative Learning Control ; Water Distriubtion Network ; AAU Lab ; Water Lab ; Leakage Detection ; Dynamic Model ; Static Model ; Artis ; Danas ; Ibrahim ; Leakage Localization ; Master Thesis ; SWIL ; Simulation ; MATLAB ; Modeling