Grouping Water-Demand Nodes by Similarity among Flow Paths in Water-Distribution Systems

Abstract

The identification of monitoring or sensor locations, or demand estimation, within drinking water distribution systems can be challenging given the size of realistic network models. Approaches such as skeletonization or aggregation can effectively reduce a network model and are generally appropriate for satisfying hydraulic objectives. However, a reduced hydraulic network model might not be appropriate for water quality objectives because of altered transport characteristics. This study proposes a clustering approach that groups nodes with similar water quality characteristics within the context of maintaining the original network structure. The proposed approach uses an input-output relationship to assess the hydraulic path between any two nodes. Using the hydraulic path information, a k-means clustering algorithm identified nodes with similar hydraulic paths. For two different case studies, as the number of clusters increased, the nodes within each cluster were shown to become more similar. The differences in water quality characteristics between the individual clusters also increased, demonstrating the ability to generate more distinct clusters of nodes. By identifying nodes with similar water quality characteristics, the resulting clusters can provide future opportunities, for example, to reduce the problem size for monitoring or sensor selection by assuming the nodes within a given cluster behave similarly.