Parallel evolutionary algorithm for Water Distribution Network Design, using the Masters–Students model in distributed environment

Abstract

The Water Distribution Networks Design Problem study is widespread in the scientific community due to its practical applicability. In this work, a sequential evolutionary algorithm has been designed, developed, and successfully applied to solve small and medium instances; it has been called the Evolutionary Algorithm for Water Distribution Network Design Problem. This algorithm is executed in centralized environments and can perfectly solve these instances in less than minutes. For large-scale real-world instances, the Evolutionary Algorithm has been adapted to work in distributed environments by using a novel parallel model, also proposed in this work, called the Masters–Students​ model. This model has been used for designing, developing, and implementing the resulting parallel evolutionary algorithm, named PEA-WDND. The Evolutionary Algorithm would last for days in the case of real-world instances, but the parallel algorithm solves them in seconds or, at maximum, in minutes. This study shows that the parallel algorithm yields an execution time lower than the execution time obtained from the evolutionary algorithm for different theoretical and practical instances.