Author(s): Nelson Carrico; Joao Caetano; Didia Covas; Soraia de Almeida
Linked Author(s): Dídia Isabel Cameira Covas
Keywords: Anomaly detection; Artificial intelligence; Data streaming; Digital water; Real-time monitoring; Water utilities
Abstract: The management of Water Supply Systems (WSS) is increasingly challenged by ageing infrastructure, water scarcity, and climate variability. While utilities collect growing volumes of operational data—from flow meters, pressure sensors, and smart meters—this information often remains underutilised due to fragmented systems and lack of integration. These challenges are particularly acute in small and mid-sized utilities, where reactive management remains the norm. The StreamWater project, addresses this gap by developing and validating a real-time, AI-enabled platform for anomaly detection in WSS. The platform processes high-frequency data streams and applies unsupervised machine learning algorithms to detect and localise anomalies such as leaks and unauthorised consumption. StreamWater is being piloted at Inframoura, the utility responsible for Vilamoura (Portugal), which features a seasonal fivefold demand variation and a rich telemetry infrastructure. The platform ingests 5-minute data from 24 flow meters, 20 pressure sensors, and 17,000 hourly smart meters. It includes modules for data pre-processing, anomaly detection, water and energy balance calculation, and operator visualisation via an integrated dashboard. Expected outcomes include a validated prototype, new open AI models for water utilities, and a methodological roadmap to support wider replication and digital transformation in the water sector.
Year: 2026