WDSEventDB: A Real-Time Benchmark Dataset for Event Diagnosis in Water Distribution Networks

Water distribution networks (WDS) are critical infrastructures whose security against stress and failures is vital to societal well-being and socio-economic development. Recent technological advances in sensor technologies have enabled the collection of real-time hydraulic datasets. These datasets can be leveraged to develop data-driven models for real-time event detection in WDS. However, most existing datasets remain proprietary and inaccessible to the public. Moreover, publicly available databases are limited in scope, as they focus on a single event type (e.g., leakages) and rely on hydraulic simulations. To address these limitations, we introduce a WDS Event Diagnosis Benchmark (WDSEventDB), a publicly available real-time sensor dataset designed to support the development of event-detection frameworks for diverse anomalies within WDS, including cyber-physical attacks, leakages, and sensor failures. The dataset is curated in real-time from a lab-scale WDS by emulating failures. We also provide four baseline detection models to facilitate benchmarking and advance research in WDS.