A Machine-Learning Approach for Monitoring Water Distribution Networks (WDNs)

The knowledge of the simultaneous nodal pressure values in a water distribution network (WDN) can favor its correct management, with advantages for both utilities and end users, guarantee higher sustainability use resource. However, monitoring all nodes is not feasible, so it be useful to develop methods that allow us estimate whole field based on data from limited number nodes. For this purpose, work employed an artificial neural (ANN) as machine-learning regression algorithm. Uncertainty demand modeled through scaling laws, linking statistics users served by each node. Three groups scenarios are generated using Latin Hypercube Random Sampling three different cross-correlations matrices demands. Each corresponding training ANN, whose performance parameter preliminarily used solve sampling design WDN. Most so-derived coincide cases. ANN appears strongly influenced cross-correlation values, best results provided relating most correlated