A Novel Framework for Efficient Hydraulic Simulations of Large Water Distribution Networks

<p dir="ltr">With urbanization and population growth, water distribution networks (WDNs) continue to expand, bringing significant challenges to the development of real-time models or digital twins for water management. Using large-scale models for multiple hydraulic simulations often results in long response times, failing to meet the demand for efficient model applications. In this study, a novel framework that combines a Joint Topology-calculation Decomposition Method (JTDM) and an Early-Stopping High-Performance Decomposed Gradient Algorithm (ES-HDGA) is proposed to accelerate hydraulic computations in large WDN models. JTDM decomposes a WDN model into a boundary model and independent sub-models, along with their hydraulic equations. ES-HDGA then uses parallel computation to solve these equations iteratively. The early-stopping strategy in ES-HDGA enables the early termination of boundary model iterations once specified accuracy is achieved, overcoming the computational bottleneck caused by boundary model computations. Tests from two cases show that ES-HDGA can reduce computation time by up to 70% compared to the widely used EPANET in a large model while maintaining acceptable accuracy, offering new insights and solutions for efficient model applications.</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><p dir="ltr"><br></p>