Nouvelle approche pour le dimensionnement et l'optimisation de la gestion des reservoirs et de la qualite de l'eau potable dans les reseaux de distribution

Water Distribution Networks (WDN) are traditionally designed without integrating tanks in the optimization process, despite the numerous benefits tanks bring to network operation. These benefits include efficient demand management, water supply in case of system failure, fire fighting or the modulation of pump flow rate. When properly designed and located, storage tanks are a cost-effective means of improving overall network performance. However, tank design is complex because several factors are to be considered with several variables involved.;This research project aims to develop a model for multi-objective optimization of water distribution systems that is able to predict the need for the introduction or not of one or more tanks in the network and the locations of these potential tanks.;In order to achieve the aim of the research, the first step consisted of a review of different approaches to including tank consideration in WDN optimization. The resulting analysis first confirms the need for tank consideration in the network design optimization process and secondly led to suggest a procedure for selecting the most suitable approach to real-world WDN optimization projects based on a number of predetermined criteria.;For the sake of proposing a model with a wide application scope, the second step consisted of providing a certain number of reference parameters that may allow the design approach generalization. These parameters are: (1) the number of pressure zones; (2) the minimum and maximum tank bottom elevations; (3) the reference diameter for pipes; (4) the reference flow and head for pumps and (5) selection of genetic algorithm parameters. They are useful for practical applications of the optimization algorithm to real practices.;The traditional approach to design tanks in distribution systems is to introduce the tank into a WDN at a predetermined location and elevation (manual selection), followed by pipe optimization, the tank volume being decided through extended-period simulation (manual check) or by guidelines. This approach allows for only a limited number of trials on tank location, elevation and storage prior to selection. It does not adequately explore the solution space and it becomes very difficult to use this approach for large networks requiring much effort before finding a feasible solution. This approach tends to implement large water tanks which are supposed hydraulically safe for water distribution systems. Those large tanks can however be a source of trouble regarding the quality of the water provided to the citizens.;In order to provide a decision support tool to network managers, the third stage was devoted to the development of the numerical model itself. A version of this model is implemented in a deterministic environment and the other in a stochastic one to allow for a comparison of both design approaches and also to give more choices to the users. The water quality was considered as part of the objectives in the network optimal design problem in both approaches.; The model was applied to three case studies in order to validate it. The results showed that including water quality in the objectives allows for an effective integration of the tanks in the optimization process. Although not being the most realistic, the deterministic approach provides solutions that can be obtained in much less time compared to the stochastic approach which may be non feasible on very large and complex networks.;Keywords : water distribution network, drinking water, pressurized flow, water tank or reservoir, tank location, simulation, design, optimization, optimal, multi-objective, decisionmaking, genetic algorithm, uncertainty, stochastic, water quality, Latin Hypercube, correlation.