Algoritmos evolucionários multipopulação na otimização multiobjetiva da remediação de águas subterrâneas

• Ano de defesa: 2009
• Autor(a) principal: Pinto, Marcos Rodrigues
• Orientador(a): Martins, Eduardo Sávio Passos Rodrigues
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Recursos hídricos; Águas subterrâneas; Contaminantes
• Link de acesso: http://www.repositorio.ufc.br/handle/riufc/16850

Through the last three decades the evolutionary algorithms have been successful on application to many areas. Easily applied, efficiency and confidence are the main advantages of the evolutionary algorithms. In the groundwater remediation, generally, the objectives are the cost minimization, minimization of contaminant presence, maximization of pumping efficiency, among others. These objectives are naturally in conflicting and the search for optimal solutions, or almost optimal solutions are needed. In view of that, the evolutionary optimization methods have been applied and refined in order to search these solutions. A brief description of these methods is presented, referring to their advantages and limitations. Five mathematical functions are used to measure the algorithms performance and allow a comparison between them. In order to optimize a “pump-and-treat” system in the remediation of a hypothetical site, multi-population evolutionary algorithms are used, considering the problem multi-objective dimension. The multi-population approach has been applied as mitigate for the main evolutionary optimization drawback: the excessive computational time. The groundwater flow modeler MODFLOW (modular finite-difference flow model) is used with the contaminant transport simulator MT3DMS (modular three-dimensional multispecies transport model). Two multi-population algorithms are presented: MINPGA (Multi-Island Niched Pareto Genetic Algorithm), that is a NPGA (Niched Pareto Genetic Algorithm) multi-population version with the injection island approach; MHBMO (Multi-Hive Honey Bee Mating Optimization), that is a HBMO (Honey Bee Mating Optimization) multi-population version. A PSO (Particle Swarm Optimization) multi-population version, called MCPSO (Multi-Swarm Cooperative Particle Swarm Optimization) is too used. Tests with mathematical functions validate the presented algorithms. Remediation problem using the “pumping-and-treat” technique had as objectives the minimization of remediation cost and minimization of contaminant final plume. The results were shown to be very good for all algorithms, but MINPGA had a tenuous advantage over others