Estudo ótimo da localização de válvulas redutoras de pressão em sistemas de distribuição de água considerando objetivos múltiplos

Abstract

In Brazil, water distribution systems with high leakage levels represent a problem with multiple implications for water utilities. Leakage, which accounts for the dominant component of water losses, produces negative effects from the economical, as well as from the operational perspectives: increase in pumping energy, compromise of water quality, increase in costs for rehabilitation or for reducing the deterioration of surrounding structures and public services. Many studies have demonstrated that leakage through cracks and orifices in a pipe increases proportionally with the increase of pressure. Therefore, optimization of a hydraulic distribution system through pressure control can lead to defining policies for effectively reducing the leakage level, the energy costs associated and thus, improving the service conditions. This study aims to develop a methodology to minimize leakage in water distribution systems by means of determining the optimal location of pressure reducing valves and following a multiobjective approach. In pursuing such purpose, optimal solutions are searched through the combination of a computer model of the system, a hydraulic simulator, EPANET 2, which has been broadly tested by the international scientific community, and a multiobjective optimization program, developed to integrate the network model and the hydraulic simulator. Two networks, named Network 1 and Network 2, were studied. In both cases, it was evidenced that introducing leakage considerably reduced the resilience of the systems. The Pareto fronts obtained in the optimization process allowed for determining the relationship between leakage, resilience and demand. By knowing those relationships, the best alternative locations for PRVs were established through the application of a decision algorithm. A manner to select the most suitable alternative was further proposed for each network. Finally, the configurations for the PRVs were determined in terms of nondimensional head loss coefficients, which can be used to select the type of PRV to be installed.