Avaliação do operador elitismo na detecção de vazamentos em rede de distribuição de água utilizando o método transiente inverso e algoritmo genético

• Ano de defesa: 2018
• Autor(a) principal: Sampaio, Amanda Sousa
• Orientador(a): Araújo, John Kenedy de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Recursos hídricos; Algoritmos genéticos; Água - Desperdício; Calibração; Inverse transient analysis; Leaks; Elitist strategy
• Link de acesso: http://www.repositorio.ufc.br/handle/riufc/32226

Detection of leakage and calibration of hydraulic models are important questions about the management of hydraulic distribution networks. When applying Genetic Algorithms to optimize calibration it is important to maintain two seemingly contradictory requirements: to preserve promising individuals from one generation to another through the elitist technique and to maintain the diversity of the population. This work evaluates the performance of the Elitismo operator used in the Genetic Algorithm (GA) search and optimization technique for locating and quantifying network leakage through the Inverse Transient Method (MTI), which consists of calibrating unknown network parameters to minimize deviations between observed and calculated transient hydraulic loads. The use of a computational application for leak calibration is demonstrated in a small hypothetical case study. The leaks were treated as additional demands on the nodes of the analyzed network, and the parameter to be calibrated is the node CDA (product of the Coefficient of Discharge by the Leak Area). The results obtained without the application of Elitism (pe = 0%) and with two other different types of Elitism were evaluated and compared, which diverge in relation to the choice of individuals that will compose the population of parameters. Based on the analyzes, the influence of Elitism on the behavior of the Objective Function, on the calculated permanent and transient loads, and on the quality of the optimization is demonstrated. The influence of the position of the casting node was also analyzed. Elitism type 2 with pe 20% was the only one to be able to identify and quantify the leakage in all the cases in which it was applied. Regardless of Elitism, the best results occurred in the node near the reservoir and in the neighboring nodes to the monitored node. It concludes with the work that Genetic Algorithm techniques are appropriate to the calibration of hydraulic networks, being a topic that deserves to be explored further.