Water utilities are challenged to reduce their water losses through detecting, localizing, and repairing leaks as quickly as possible in their aging distribution systems. In this work, we solve this challenging problem by detecting multiple leaks simultaneously in a water distribution network for the Battle of the Leak Detection and Isolation Methods. The performance of leak detection and localization depends on how well the system roughness and demand are calibrated. In addition, existing leaks affect the diagnosis performance unless they are identified and explicitly represented in the model. To circumvent this chicken-and-egg dilemma, we decompose the problem into multiple levels of decision-making (a hierarchical approach) where we iteratively improve the water distribution network model and so are able to solve the multileak diagnosis problem. First, a combination of time series and cluster analysis is used on smart meter data to build patterns for demand models. Second, point and interval estimates of pipe roughnesses are retrieved using least squares to calibrate the hydraulic model, utilizing the demand models from the first step. Finally, the calibrated primal model is transformed into a dual model that intrinsically combines sensor data and network hydraulics. This dual model automatically converts small pressure deviations caused by leaks into sharp and localized signals in the form of virtual leak flows. Analytical derivations of sensitivities with respect to these virtual leak flows are calculated and used to estimate the leakage impulse responses at candidate nodes. Subsequently, we use the dual network to (1) detect the start time of the leaks, and (2) compute the Pearson correlation of pressure residuals, which allows further localization of leaks. This novel dual modeling approach resulted in the highest true-positive rates for leak isolation among all participating teams in the competition.

@en

In this manuscript, we consider the problem of optimally scheduling the restoration of a water distribution network with multiple failures after a disaster. The decisions made are the sequence of burst/broken pipes and leaks that need to be replaced or repaired, respectively, subject to constraints on workforce availability and physical hydraulic conditions. In order to sufficiently capture the objectives of the utility (e.g. service levels, resilience loss and customer minutes lost without service), which are all pressure dependent, pressure driven leakage and demand models (PDM) are employed.
The restoration decisions are modelled as time-dependent closure and opening of links and are simulated using a PDM in EPANET, propagating decisions as pressure-driven hydraulic constraints and computing a posteriori their impact on the multiple desired restoration objectives, some of which have trade-offs. The combination of discrete decisions with time-dependent couplings, and the presence of objectives that are conditional functions of demands met and time indices make it difficult to pose this optimal task scheduling problem as a standard numerically tractable mixed-integer programming problem. To make the problem tractable for the given large-scale water distribution system, we propose greedy heuristics that use a hierarchical decomposition of the decision space using structural properties of the network graph and hydraulics. Firstly, PDM simulations are used to sort the breaks and leaks from the biggest losses to the smallest, or determine visibility of the damages. In addition to solving the multi-criteria scheduling problem, we also use engineering principles to derive metaheuristic that can prioritise water loss reductions. The greedy algorithm is employed to iteratively schedule isolation and repairs by first stabilizing the system with the isolation of the biggest breaks; an alternative approach considers all objectives ‘equally’. With these we explore the trade-offs in response between water loss and resilience indicators.
To enable the scheduling, we use graph decomposition techniques to identify the valves that need to be closed to isolate a hydraulic segment (i.e. set of links sharing same closing valves) for replacement; this gives us a map (or look up table) that will be used in the scheduling. The map also includes information about the number of nodes isolated, unsatisfied demand when isolating each segment and the total pipe length of the segments. We also analyse the system flows, network pressures and how the depletion of tanks affects service levels. Using these, we make recommendations for improving the capacity of the system, including the improvement of pumping stations, installation of control valves and some pipe re-enforcement. The same greedy task scheduling algorithm is then used under these alternative network improvements, to show a much better response in all criteria.@en

This paper considers the sectorization of a large water distribution network into district metered areas (DMAs) and simultaneously optimizes rehabilitation of the network with new pipes, control valves, and storage tanks. Since the available water resources are much smaller in the dry season, both the design and operational settings are optimized to satisfy water demand, water quality and pressure constraints, and efficiency indices under stringent conditions. Because of the heterogeneity of the multiple decision variables and the complicated way they interact through the multiple objectives (some complimentary and some conflicting), it is not possible to fully automize the simultaneous sectorization, rehabilitation and operational optimization. Therefore, we employ a multi-stage approach where engineering judgement and network graph simplification and visualization tools are employed to find a good feasible solution that is used as a first guess for further optimization of sectors and operational settings, to achieve feasible solutions with better cost of implementation, demand similarity among DMAs and better pressure uniformity in operations. A multi-objective Agent Swarm Optimization framework is used to iteratively change the sectors at the boundaries. For the final configuration, sequential linear programming is used to find optimal valve and pump settings. @en