Data Assimilation in Water Distribution Systems

Operational management of water distribution networks (WDNs) requires assimilating observations such as pressures at junction nodes, flows in pipes and, whenever available, monitored demand. Although several data assimilation techniques are today available, ranging from 1-2-3-4 DVAR to Kalman Filters, a problem is posed by the need of preserving the structural relations among state variables in a WDN, such as for instance pressure head, discharge and demand. Ensemble Kalman Filters certainly can be used to account for non-linearities but, for example, if one tries to assimilate pressure heads and pipe flows at the same time, nothing guarantees that the resulting variables after the data assimilation step, will still obey to the hydraulic structural relations mathematically describing a WDN. In this work an EnKF based procedure has been implemented, which allows to assimilate three types of observations, namely pressures at junction nodes, flows in pipes and monitored demand. The procedure allows to assimilating all the observations in three successive steps, while guaranteeing the full satisfaction of the structural relations. The results, demonstrated over an operational network, show the high performance of the chosen approach.