Multistage iterative fully automatic partitioning in water distribution systems

Abstract

This paper presents a novel method using a clustering, detection, and optimization model to devise a solution of fully automatic partitioning in a water distribution system (WDS). First, the Black Hole Clustering Algorithm is employed to divide the WDS into different partitions. Second, two types of outliers are eliminated by multistage iterative processes including traverse, k-Nearest neighbor, and the Warshall algorithm. Finally, the boundary conditions of the partitions are optimized by a Non-dominated Sorting Porcellio Scaber Algorithm to minimize the number of boundary pipes required to balance pressures and reduce leakages. Seven WDSs are employed as case studies to verify the practicability of the method. The Open Water Analytics toolbox is applied to code the hydraulic calculation program. The result demonstrates that average pressure and leakage cost decreases after optimization.

Keywords

• black hole clustering
• multi-objective optimization
• network partitioning
• outlier detection