Publication

Leak learning for graph-based state interpolation in water distribution networks

Conference Article

Conference

European Control Conference (ECC)

Edition

2024

Pages

110-115

Doc link

https://doi.org/10.23919/ECC64448.2024.10591097

File

Download the digital copy of the doc pdf document

Abstract

Graph-based state interpolation (GSI) is a state-of-the-art state reconstruction technique that operates over water distribution networks (WDN). This method retrieves the complete hydraulic state (nodal heads) of the network based on its topology and limited pressure measurements. To perform leak localization, GSI is coupled with a process that compares interpolated leak and leak-free states. This article presents a methodology to adapt GSI in order to learn from its off-line and online operation (i.e., gain knowledge about historical located leaks, as well as leaks appearing in the network) and using this information to improve localization in future leak events. The methodology is tested over a well-known case study (Modena), showing promising results in terms of localization performance.

Categories

learning (artificial intelligence), nonlinear programming, optimisation.

Author keywords

leak localization, water distribution network, learning, state interpolation

Scientific reference

L. Romero, G. Cembrano and V. Puig. Leak learning for graph-based state interpolation in water distribution networks, 2024 European Control Conference, 2024, Stockholm (Sweden), pp. 110-115.