Comparative study of multi-objective evolutionary algorithms for hydraulic rehabilitation of urban drainage networks

Abstract

Multi-Objective Evolutionary Algorithms (MOEAs) are flexible and powerful tools for solving a wide variety of non-linear and non-convex problems in water resources engineering contexts. In this work, two well-known MOEAs, the Strength Pareto Evolutionary Algorithm (SPEA2) and Non-dominated Sorting Genetic Algorithm (NSGA2), and two additional MOEAs that are extended versions of harmony search (HS) and differential evolution (DE), are linked to the Environmental Protection Agency's Storm Water Management Model (SWMM-EPA), which is a hydraulic model used to determine the best pipe replacements in a set of sewer pipe networks to decrease urban flooding overflows. The performance of the algorithms is compared for several comparative metrics. The results show that the algorithms exhibit different behaviours in solving the hydraulic rehabilitation problem. In particular, the multi-objective version of the HS algorithm provides better optimal solutions and clearly outperforms the other algorithms for this type of nondeterministic polynomial-time hard (NP-hard) problem.