Author(s): Raluca Iulia Cozma Mircescu; Gabriel Viscarret Atienza; Humberto Bustince Sola; Francisco Javier Fernandez Fernandez
Linked Author(s):
Keywords: Anomaly detection; Matrix profile; Time-series clustering; Water distribution
Abstract: Non-revenue water (NRW), primarily lost through leaks and pipe bursts, represents a global economic and environmental challenge with daily losses of 45 million m3 and over USD 3 billion annually. Reducing these losses by half could supply water to approximately 90 million people [1]. Early detection and prediction of anomalies are therefore crucial to improving efficiency, sustainability and infrastructure resilience. Recent advances in deep learning and digital twin technologies enable real-time leak detection, but most existing approaches still lack continual learning mechanisms, predictive capabilities and efficient data fusion strategies.
Year: 2026