Leak Localization Method for Water-Distribution Networks Using a Data-Driven Model and Dempster-Shafer Reasoning

This article presents a new data-driven method for leak localization in water-distribution networks (WDNs). The method uses the information provided by a set of pressure sensors installed in some internal network nodes in addition to flow and pressure measurements from inlet nodes. Pressure measurem...

Descripción completa

Detalles Bibliográficos
Autores: Soldevila, Adrià, Blesa, Joaquim, Jensen, Tom Nørgaard, Tornil-Sin, Sebastian, Fernández Cantí, Rosa M., Puig, Vicenç
Tipo de recurso: artículo
Estado:Versión aceptada para publicación
Fecha de publicación:2021
País:España
Institución:Consejo Superior de Investigaciones Científicas (CSIC)
Repositorio:DIGITAL.CSIC. Repositorio Institucional del CSIC
OAI Identifier:oai:digital.csic.es:10261/261219
Acceso en línea:http://hdl.handle.net/10261/261219
Access Level:acceso abierto
Palabra clave:Data models
Fault diagnosis
Leak detection
Water resources
Descripción
Sumario:This article presents a new data-driven method for leak localization in water-distribution networks (WDNs). The method uses the information provided by a set of pressure sensors installed in some internal network nodes in addition to flow and pressure measurements from inlet nodes. Pressure measurements are recorded under leak-free network operation, and a WDN data-driven model of the pressure at each sensed node is adjusted. The pressure estimation from this model is complemented by a Kriging spatial interpolation technique to estimate the pressure in the nodes that are not sensed, leading to a pressure reference map. Leak localization is based on the comparison of this reference pressure map with the current pressure map that is obtained by applying Kriging directly to the pressure measurements provided by sensors. The key element in this comparison is the use of the Dempster-Shafer theory for reasoning under uncertainty. The successful application of the proposed methodology to two real-data case studies is presented.