Author(s): Stefania Piazza; Gabriele Freni
Linked Author(s): Gabriele Freni
Keywords: Leak detection; Physically-based leak models; Water distribution networks
Abstract: Identifying and reducing water losses in Water Distribution Networks (WDN) is a key priority for utilities, especially in light of increasing energy costs and sustainability goals. In Italy, water losses are assessed through the macro-indicators M1b, M2 and M3 defined by ARERA, with M1b (percentage of lost volume on distributed volume) as a key benchmark for performance improvement and investment planning. This study aims to develop a reliable simulation model for WDNs to support asset management strategies focused on “smart” pipe replacement, focusing only on the most critical network segments. To do so, a hybrid methodology combining physics-based hydraulic models with AI-based data analytics has been proposed to improve reliability and adaptability. Several Machine Learning (ML) and Artificial Intelligence (AI) methods have been applied to locate water leaks. The developed methodology has been applied to the Laboratory network of the University of Enna “Kore”. The results demonstrate the effectiveness of the model in detecting water leaks in the presence of limited information.
Year: 2026