Benchmarking Multi-Objective Optimizers for Climate-Stressed Urban Drainage Network Retrofit Planning: A Simulation-Budget Comparison

Sewer retrofits for a wetter climate are increasingly planned with simulation-based multi-objective optimization, but each candidate plan must be evaluated with a hydrodynamic model, so only a few thousand simulations can be afforded, and which optimizer uses them best is not known. This study compares eleven search strategies—three established Pareto-dominance-based evolutionary algorithms, four algorithms designed for many objectives, a particle-swarm and a differential-evolution algorithm, random search, and a local search—on one problem: adding up to three relief conduits to the calibrated SWMM model of the Banpo trunk–sewer catchment in Seoul to minimize cost and the worst remaining node flooding and to maximize the flood-volume reduction under a 100-year storm intensified by 30%, with 2000 simulations per run. The dominance-based algorithms performed best and almost equally well, the many-objective algorithms brought no benefit, and two methods fell below random search; the population size, however, changed the result of NSGA-II by more than the differences among the leading algorithms, so the ranking is conditional on the shared configuration. One optimizer run for the full budget used a fixed number of simulations better than combinations of shorter runs of several optimizers. Most of the 869 reference plans shift part of the flooding to the receiving trunk junctions, which the objectives do not reveal, whereas their ranking was insensitive to 14 tested rainfall intensities, patterns, and durations. For practice, the results support one established optimizer, the full run length, several seeds, and node-level verification; for benchmarking, they show that the candidate screening and the algorithm configuration, rather than the choice among established optimizers, bound what can be concluded.

Authors

Institutions

Publication Details

Journal
Water
Published
2026-11-16
DOI
https://doi.org/10.3390/w18222859
Primary Topic
Urban Stormwater Management Solutions
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Benchmarking Multi-Objective Optimizers for Climate-Stressed Urban Drainage Network Retrofit Planning: A Simulation-Budget Comparison

Jaekyoung Kim, Samuel Park
Water
Urban Stormwater Management Solutions
article

Benchmarking Multi-Objective Optimizers for Climate-Stressed Urban Drainage Network Retrofit Planning: A Simulation-Budget Comparison

Jaekyoung Kim, Samuel Park
article en

Abstract

Sewer retrofits for a wetter climate are increasingly planned with simulation-based multi-objective optimization, but each candidate plan must be evaluated with a hydrodynamic model, so only a few thousand simulations can be afforded, and which optimizer uses them best is not known. This study compares eleven search strategies—three established Pareto-dominance-based evolutionary algorithms, four algorithms designed for many objectives, a particle-swarm and a differential-evolution algorithm, random search, and a local search—on one problem: adding up to three relief conduits to the calibrated SWMM model of the Banpo trunk–sewer catchment in Seoul to minimize cost and the worst remaining node flooding and to maximize the flood-volume reduction under a 100-year storm intensified by 30%, with 2000 simulations per run. The dominance-based algorithms performed best and almost equally well, the many-objective algorithms brought no benefit, and two methods fell below random search; the population size, however, changed the result of NSGA-II by more than the differences among the leading algorithms, so the ranking is conditional on the shared configuration. One optimizer run for the full budget used a fixed number of simulations better than combinations of shorter runs of several optimizers. Most of the 869 reference plans shift part of the flooding to the receiving trunk junctions, which the objectives do not reveal, whereas their ranking was insensitive to 14 tested rainfall intensities, patterns, and durations. For practice, the results support one established optimizer, the full run length, several seeds, and node-level verification; for benchmarking, they show that the candidate screening and the algorithm configuration, rather than the choice among established optimizers, bound what can be concluded.

WaterVol. 18(22)
University of Seoul (KR)
National Research Foundation of Korea, Korea Agency for Infrastructure Technology Advancement
Sustainable cities and communities, Clean water and sanitation, Climate action
Urban Stormwater Management Solutions
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon