Optimisation of large-scale water distribution system design using genetic algorithms

Abstract

This paper presents details of a relatively new approach to the design of water distribution systems. An optimization technique called genetic algorithms is used to evolve an improving population of designs. Only a very, very small proportion of the total possible search space needs to be explored to find low cost solutions that satisfy all of the specified design criteria. The genetic algorithm (GA) uses operators of selection, crossover and mutation applied to a set of strings that represent the decision variables to be selected as part of the design process. These operators enable the genetic algorithm process to quickly seek out low cost optimal solutions. Many tests of the application of the genetic algorithm optimization process to real-life network designs has shown that the GA is effective at finding low cost solutions. The technique has consistently found lower cost solutions than the trial-and-error simulation approach typically used by design engineers. In addition, the technique is more effective and easier to apply than traditional mathematical optimization techniques.