PipeNetGen: A methodology for generating synthetic water distribution network models

Water distribution networks (WDNs) are centralized systems that deliver water across large geographic areas to diverse user groups. WDN models are critical in developing designs, operations, and management strategies for water infrastructure systems and in the research and development of new methodologies to support decision-making. WDN models, however, are time consuming to create, and under-resourced water utilities may not have the resources to develop a WDN model. This research develops a new method for generating synthetic WDN models using open-source data, graph theory, and Mixed-Integer Linear Programming (MILP). This research introduces the Pipe Network Generator methodology (PipeNetGen), which can create a synthetic WDN model for any city using open-source data or private utility information. PipeNetGen includes a core MILP model based on previous research and combines it with a clustering algorithm to run larger systems, a hydraulic simulator to correct MILP assumptions, and a fire flow method to ensure network reliability under high demands. PipeNetGen is applied to Lakewood, California, and the synthetic WDN model is compared with a water utility-provided engineered WDN model based on pipe sizes, topology, hydraulics, and demand. PipeNetGen provides a valuable resource to improve water infrastructure management for utilities and research.