Dimensionamento econômico de redes de distribuição de água aplicando algoritmo genético

Abstract

A computer model for least-cost design of water distribution network was developed in two modules: hydraulic simulation by EPANET-2 and multi-objective optimization through Genetic algorithm GA. Sensitivity analyses were done trying individually each parameter getting values that improve the algorithm performance in convergence, quality of solutions and computer effort. Using these parameters values, were done longer simulations to find the best solution to the first objective function that approaches the section pipes efficiency. Its result is compared to the nonlinear programming one, got for the same water network, presenting a more efficient network hydraulically and so cheaper than the last. It was implemented the second objective function, now approaching the total least-cost, of network construction plus pumping costs, comparing to the first objective function results. The second objective function utilization is much more advantageous. Although it is not assured the best solution found is the global optimum, the GA method for least-cost design of water network performs perfectly feasible face conventional optimization techniques, and the hybridization of such methods must provide convergence velocity and accurate results.