Sectorization of a Water Distribution Network (WDN) into District Metered Areas (DMAs) is a proven solution for proactive leakage control. Traditionally, WDN sectorization is done using a "trial and error" approach conducted by local experts which often results in arbitrary solutions being identified. A number of methods published recently tried to improve WDN sectorization by automating the process, especially by using optimization. Various sectorization criteria, constraints and limitations are introduced, often neglecting limited funds and shortage of water balance data often encountered in poorly managed WDNs. These methods also suffer from low computational efficiency imposed by optimization methods used. This paper presents a new, Distribution Network SEctorization (DeNSE) method that overcomes these deficiencies. The new method is based on a heuristic procedure where the WDN sectorization is driven by efficient tracking of water balance and least cost investment for implementation while maintaining the same level of WDN's operational performance. Aforementioned set of criteria is particularly well suited for initial sectorization of poorly managed WDNs, in which great uncertainty in water balance data often leads to poor management decisions. DeNSE method is validated and benchmarked against several literature sectorization methodologies on a real-sized WDN. The results obtained demonstrate the ability of the DeNSE to identify set of good, realistic sectorization solutions that are in some respects better than the corresponding solutions reported in the literature. The new method also enables sectorization to be done in a computationally efficient manner ensuring its applicability to large, real-life sized WDNs.

No abstract

Notwithstanding recent advances in hydrological modelling, flood simulations remain challenging since many processes must be simulated with high computational efficiency. This paper presents a novel geographic information system (GIS)-oriented platform 3DNet and the associated hydrologic model, with focus on the platform and model features that are relevant for flood simulations. The platform enables hydraulic structures to be incorporated in the hydrologic model, as well as water retention. A limiting capacity can be imposed on every river reach enabling estimation of flooding volume. Runoff is simulated within irregularly shaped units that can be aggregated providing spatial flexibility, i.e. model setup can vary from lumped to semi- and fully-distributed. The model contains many parameters with a physical connotation that can be inferred from catchment characteristics, and it enables simulations with minimum data requirements. All algorithms are implemented in C++ warranting fast computations, while the spatial flexibility can provide additional speed-up. The model is used for a reconstruction of a devastating flood in the Kolubara catchment in May 2014. Despite incomplete and uncertain observations, reasonable results across the catchment are obtained with the plausible parameter estimates. The results suggest that enclosure of the presented features in flood simulation tools would improve simulation accuracy and efficiency.

To optimize the design of a water distribution network (WDN), a large number of possible solutions need to be examined; hence computation efficiency is an important issue. To accelerate the computation, one can use more powerful computers, parallel computing systems with adapted hydraulic solvers, hybrid algorithms, more efficient hydraulic methods or any combination of these techniques. This paper explores the possibility to speed up optimization using variations of the ΔQ method to solve the network hydraulics. First, the ΔQ method was used inside the evaluation function where each tested alternative was hydraulically solved and ranked. Then, the convergence criterion was relived in order to reduce the computation time. Although the accuracy of the obtained hydraulic results was reduced, these were feasible and interesting solutions. Another modification was tested, where the ΔQ method was used just once to solve the hydraulics of the initial network, and the unknown flow corrections were added to the list of other unknown variables subject to optimization. Two case networks were used for testing and were compared to the results obtained

No abstract