Boosting Genetic Algorithm Performance in Pump Scheduling Problems with a Novel Decision-Variable Representation

Abstract

Pumping stations used in water distribution networks (WDNs) consume a significant portion of the energy required to deliver municipal drinking water. Smart management strategies such as optimal pump scheduling (OPS) have gained the attention of water companies and managing authorities because they help reduce both energy costs and detrimental consequences for the environment. Genetic algorithms (GAs) are frequently used to approximate the solution of OPS problems, although many researchers have resorted to hybrid models to improve computational performance. This paper shows that despite the lack of support in the literature, a well-designed GA is capable of tackling OPS problems effortlessly. In addition, a new decision-variable representation is proposed, specifically suited to parallel pump systems, which is able to further improve the performance of a GA. Finally, the outperforming capabilities of the new variable representation are demonstrated with two case studies from recent literature.