Estudo da influência do operador cruzamento na calibração do fator de atrito em rede de distribuição de água utilizando o método do transiente inverso acoplado à técnica de algoritmo genético

• Ano de defesa: 2019
• Autor(a) principal: Pinto, Átila Alves
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Recursos hídricos; Calibração; Transitórios hidráulicos; Algoritmo genético; Friction factor; Inverse transient; Crossover operator
• Link de acesso: http://www.repositorio.ufc.br/handle/riufc/48226

Mathematical models for hydraulic optimization and simulation have been developed, with high investments from the interest of water companies, aiming at the efficient operation of their water supply systems. Ifthere are significant discrepancies between measured and modeled datamodel, calibration becomes critical. Optimization techniques with hydraulic simulators have been used to estimate values for unknown hydraulic parameters of a network, for example Genetic Algorithms, based on the biological theory of evolution, canbepointedout.The influence of the crossoveroperatorwas investigated, using the Genetic Algorithm optimization technique with the Inverse Transient Method, aiming the calibration of the friction factors of a hypothetical water distribution network. The research consists of calculating the transient and permanent hydraulic headsof the water network nodes, based on the Characteristics Method and evaluating the performance of the optimal solutions of the objective functions for the followingtypes of crossovers: arithmetic, mean, BLX-α and heuristic.The hydraulic transient was caused by a variation of the demand in monitored node (4), which lasted 30 seconds, in time intervals of 0,3seconds.The number of generations will be considered as the stopping criteriaof the genetic algorithm employed.Source codes, based on genetic algorithms, were developed to identify optimal hydraulic load values on the monitored node,and the calibration efficiency is measured by Error Relative, Mean ErrorRelativeand significance percentage by the objective function to the lowest relativeerror. The results showed that the arithmetic and mean crossoverswere pointed as the most efficient in identifying the friction factors; for the estimation of permanent and transient hydraulic loads, the BLX-α and heuristic crossoversstood out as the most effective. Therefore, it was found that there is no convergence of responses for a single type of crossover. Based on the analysisperformed, it was concluded that the REand MERindicators do notalways lead to the best results in the estimation of friction factors and hydraulic loads, as often the MERcan be considerably low, but there are high percentages of REs involved. Therefore, it is also necessary to verify the percentage of significance throughof the objective function for the smallest RE, that quantifiesthe number of generations in which the RE is minimal.