Optimal rehabilitation for large water distribution systems using genetic algorithms

Abstract

Finding the optimal rehabilitation strategy for upgrading the performance of existing water networks to meet an increasing water supply demand is an active research field. Genetic algorithms have been employed to optimise the rehabilitation of water distribution systems (WDSs) for the last two decades. However, there is a lack of methods for assessing the search performance of genetic algorithms applied to large optimisation problems as the global optimal solution for large problems is typically unknown. This paper aims to investigate the search ability of genetic algorithms applied to a number of case studies with the number of decision variables ranging from 21 to 1050 pipes. The performance assessment of genetic algorithms is made by comparing the optimal solution found by genetic algorithms and the estimated global optimal solution for each case study. It can be seen that, from the results obtained, genetic algorithms exhibit a high overall search performance when the number of decision variables is less than 100. However, the performance of genetic algorithms deteriorates when dealing with relatively larger WDS case studies with number of decision variables more than 200.