EUROPT 2025
Abstract Submission

499. Efficient Design of Water Distribution Networks: An Approximation and a Heuristic

Invited abstract in session WC-12: Optimisation under uncertainty for sustainability, stream Applications: AI, uncertainty management and sustainability.

Wednesday, 14:00-16:00
Room: B100/8009

Authors (first author is the speaker)

1. Nitish Dumoliya
Department of Industrial Engineering and Operations Research, Indian Institute of Technology Bombay, India
2. Ashutosh Mahajan
IEOR Department, IIT Bombay

Abstract

A novel approximation of the Hazen-Williams equation and an heuristic are proposed to design cost-effective water distribution networks (WDNs). WDN design is challenging in cyclic networks due to nonlinear, nonconvex headloss constraints. Pipe size selection is limited to commercially available diameters. WDNs design model optimizes pipe sizes, lengths, flows, and node heads to minimize cost while meeting demand and hydraulic constraints. WDN models have two main challenges. First, some solvers abort, flagging an undefined first derivative due to headloss equation formulation, even though it is mathematically well defined. Second, the headloss equation's second derivative is undefined at zero flow. We develop an approximation of the Hazen-Williams equation, ensuring differentiability. We compare the behaviour of benchmark instances on solvers and modeling languages to highlight these challenges. Comparative results are given for the approximate model. We also propose a heuristic based on the acyclic flows in the water network. The idea is to solve a sequence of nonlinear optimization models to find locally optimal solutions. In each iteration, flow direction is fixed in some carefully selected arcs based on the last local solution of the nonconvex model, and it allows us to explore promising locally optimal solutions that may be close to the global solution. We compare the performance of our heuristic on benchmark water network instances to global and local solvers.

Keywords

Status: accepted

Back to the list of papers