Route Delineation of Water and Irrigation Networks Using Evolutionary Optimization

<p dir="ltr">This paper presents an optimization framework for designing cost-effective and hydraulically efficient layouts and pipe diameters in water distribution networks (WDNs) using a combination of Genetic Algorithms (GA), Steiner tree theory, and linear programming. The approach minimizes total network cost while satisfying hydraulic feasibility, pressure, and flow constraints. It integrates spatial routing with hydraulic analysis, offering a robust tool for urban water infrastructure planning. A grid-based Steiner tree is used to generate the backbone of the WDN, allowing for optimal routing via intermediate nodes. The network is modeled as a graph, and a custom GA—featuring domain-specific selection, crossover, and mutation operations—is used to evolve efficient layouts. A linear programming model optimizes pipe diameters based on hydraulic constraints, with head losses computed using the Hazen-Williams equation. The framework is implemented in Python, utilizing NetworkX for graph operations, PySCIPOpt for pipe diameter optimization, and incorporates parallel processing for scalability. It also supports visualization, DXF export for CAD use, and detailed reporting on pipe sizes, flows, and hydraulic grades. Innovations include integrating Steiner tree theory with hydraulic design, adaptive GA for improved convergence, and the ability to manage large-scale networks through grid refinement. A case study demonstrates significant cost savings and faster design times—hours versus days—while meeting all hydraulic requirements. The scalable framework can be adapted for diverse urban planning scenarios. Future extensions will address multi-objective optimization (e.g., water quality, reliability) and integration with real-time data for dynamic network adaptation.</p><p dir="ltr">This paper was presented at the 21st Computing and Control in the Water Industry Conference (CCWI 2025) at the University of Sheffield (1st - 3rd September 2025).</p>