Multiobjective Water Distribution Network Planning: Hybrid ML-NSGAII-TOPSIS Approach

The planning and management of water distribution networks (WDNs) has become increasingly challenging due to rapid urbanization, aging infrastructure, and climate-induced uncertainties. This study introduces a novel modeling framework that integrates machine learning (ML) surrogate models with the Non-dominated Sorting Genetic Algorithm II (NSGA-II) to efficiently tackle WDN planning challenges. Using benchmark WDNs GoYang and D-Town, the modeling framework aimed to minimize WDNs’ life-cycle cost (LCC), reduce greenhouse gas (GHG) emissions, and maximize resilience index (RI) in a multiobjective manner considering various growth scenarios, including High Growth, Peripheral Densification, and Infill Development. By leveraging ML surrogates, the framework significantly reduces computational demands while preserving accuracy, enabling rapid assessment of potential solutions without skeletonization. The Pareto-optimal solutions derived from NSGA-II are ranked using the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) to identify the best-balanced configurations. Results show that least-cost optimization alone can compromise resilience and environmental outcomes, whereas modest additional investments often delivered substantial gains, underscoring the value of a multiobjective approach. The findings highlight the importance of balancing cost, resilience, and environmental impacts in WDN planning. The integration of ML with multiobjective optimization not only enhances computational efficiency but also equips urban planners and engineers with powerful, data-driven insights, ultimately enabling the development of resilient, sustainable infrastructure to meet the demands of an uncertain future.