Data-driven leak detection and localization using LPWAN and Deep Learning

• Autor(a) principal: Rolle, Rodrigo P. [UNESP]
• Data de Publicação: 2022
• Outros Autores: Monteiro, Lucas N. [UNESP], Tomazini, Lucas R. [UNESP], Godoy, Eduardo P. [UNESP]
• Tipo de documento: Artigo de conferência
• Idioma: eng
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://dx.doi.org/10.1109/MetroInd4.0IoT54413.2022.9831619; http://hdl.handle.net/11449/240663
• topic: Deep Learning; Graph Neural Networks; Internet of Things; Leak detection; Smart Cities

Management of water resources is a big challenge that draws the attention of global initiatives such as the Sustainable Development Objectives of the United Nations. The technological paradigm of the Internet of Things (IoT) provides the potential to enable Smart Cities, which emphasize rational consumption and waste reduction. This work proposes a system to monitor and identify leakages on Water Distribution Networks (WDNs). The monitoring devices must operate in Low-Power Wide Area Networks (LPWAN), networks that enable low power consumption at the cost of limited data throughput. A case study WDN was created on a software environment for data collection in various operation scenarios, including leakages in different locations. The obtained data sets were analyzed through data inference techniques to identify separable classes or features. Then, a Deep Learning algorithm was used to estimate the probable location of leaks in the WDN. The results obtained in the proposed case study indicate that the Deep Learning approach is a viable methodology to identify and locate leakages, despite the limited data throughput from LPWAN technologies.