Data-driven Leak Localization in Water Distribution Networks via Dictionary Learning and Graph-based Interpolation

TL;DR

This paper introduces a novel data-driven leak localization method for water distribution networks that combines graph-based interpolation with dictionary learning, improving accuracy and resilience over individual approaches.

Contribution

The paper presents a new integrated approach combining graph interpolation and dictionary learning for leak localization in WDNs, demonstrating superior performance.

Findings

Outperforms individual methods in leak localization accuracy

Validated on the L-TOWN benchmark dataset

Enhances resilience to interpolation errors and high dimensionality

Abstract

In this paper, we propose a data-driven leak localization method for water distribution networks (WDNs) which combines two complementary approaches: graph-based interpolation and dictionary classification. The former estimates the complete WDN hydraulic state (i.e., hydraulic heads) from real measurements at certain nodes and the network graph. Then, these actual measurements, together with a subset of valuable estimated states, are used to feed and train the dictionary learning scheme. Thus, the meshing of these two methods is explored, showing that its performance is superior to either approach alone, even deriving different mechanisms to increase its resilience to classical problems (e.g., dimensionality, interpolation errors, etc.). The approach is validated using the L-TOWN benchmark proposed at BattLeDIM2020.