Pipe Cost Optimization in Water Distribution Network: An Analysis of Different Constraint Handling Methods

Evolutionary algorithms (EAs) have been used to solve Water Distribution Network (WDN) Optimization problems for several decades, and several ways for resolving constrained problems have been devised. According to the no-free lunch theorem, different CHTs would perform differently on different problems due to differences in landscape, dimensionality, and ratio between the feasible search space and the entire search space. However, in WDN no work isolated the performance of different CHTs in the literature. Therefore in this work, an analysis of four well-known CHTs is presented based on 5 different case studies of WDNs. Specifically, the performance of the CHTs is analyzed in terms of pipe cost optimization of WDNs with Self-adapting differential evolution (SADE) as the core search algorithm. Results from the analysis show that certain CHTS are not suitable for WDNs and motivated the need to develop advanced variants of some CHTs to ensure their scalability to larger problem dimensions and also improve their convergence speed.