Optimization of water distribution system using improved genetic algorithms

Abstract

During the past two decades, much engineering attention has been focussed on the development for constructing optimization models and algorithms to minimize the cost of the entire water supply system. And over the last 5 year a methodology for the application of genetic algorithms to water distribution system optimization has been developed. Genetic algorithms have many advantages to describe and optimize the water distribution system more realistic than previous optimization methods. And the performance of conventional simple genetic algorithms can be improved by the change of coding scheme or adoption of additional genetic operators. So in this study the optimization of water distribution system using improved genetic algorithms is studied. The improvement of genetic algorithms is accomplished by the variation of procedure of simple genetic algorithms. Some improvements are found in other engineering fields, and others are developed for this study. And the genetic algorithms applied these improvements are simulated in a given network, and results are compared with those of simple genetic algorithms. And the improvements affecting the performance of genetic algorithms seriously are found, and finally the new and highly improved genetic algorithms are developed by the combination of these improvements simultaneously. Two example networks are simulated for confirmation of new algorithms performance. The first example considers the problem of optionally expanding an existing network, namely, New York water distribution system, and the second of the problems deals with the design of a completely new network for the city of Hanoi, Vietnam. New algorithms converge to optimal solution faster than conventional genetic algorithms in both cases, and in addition more reduced result is found in case of Hanoi water distribution system. The promotion of competition and randomization between chromosomes are both accomplished, and the optimization of algorithms themselv...