Use of Domain Knowledge to Increase the Convergence Rate of Evolutionary Algorithms for Optimizing the Cost and Resilience of Water Distribution Systems

Abstract

Evolutionary algorithms (EAs) have been used extensively for the optimization of water distribution systems (WDSs) over the last two decades. However, computational efficiency can be a problem, especially when EAs are applied to complex problems that have multiple competing objectives. In order to address this issue, there has been a move toward developing EAs that identify near-optimal solutions within acceptable computational budgets, rather than necessarily identifying globally optimal solutions. This paper contributes to this work by developing and testing a method for identifying high-quality initial populations for multiobjective EAs (MOEAs) for WDS design problems aimed at minimizing cost and maximizing network resilience. This is achieved by considering the relationship between pipe size and distance to the source(s) of water, as well as the relationship between flow velocities and network resilience. The benefit of using the proposed approach compared with randomly generating initial populations in relation to finding near-optimal solutions more efficiently is tested on five WDS optimization case studies of varying complexity with two different MOEAs. The results indicate that there are considerable benefits in using the proposed initialization method in terms of being able to identify near-optimal solutions more quickly. These benefits are independent of MOEA type and are more pronounced for larger problems and smaller computational budgets.