Optimum design of a water distribution system (WDS) by applying the Simulated Annealing (SA) Algorithm

Abstract

In this paper, an application of the Simulated Annealing (SA) algorithm for the optimum design of a water distribution system (WDS) is presented. The SA is a metaheuristic search algorithm, based on the analogy between the physical process of annealing in metals (controlled cooling of a heated mass) and the resolution of combinatorial optimisation problems. The SA algorithm, along with mathematical models, has successfully been used in the optimum design of WDS. The real-scale Marsella WDS, a rural community located in San Carlos, Costa Rica was selected as case study. The SA algorithm was implemented in the well-known open-source model EPANET through the WaterNetGen extension. Three different automated variations of the SA algorithm were compared against the man-made trial-and-error design of the Marsella WDS using pipe diameters unitary costs only. Results show that all three automated SA approaches return unitary costs under 0.49 the original cost of the trial-and-error approach as fraction. This proves that the SA algorithm is able to handle the combinatorial optimization problem of the least cost design of water distribution systems at real-scale.