Water distribution system classification using system characteristics and graph-theory metrics

This paper proposes a system-characteristics and graph theory-based water distribution system (WDS) model classification scheme that is based on system function and topology. Various parameters are examined to determine the most adequate parameter(s) for describing a WDS. The classification scheme is applied to a single hypothetical and 25 real systems. The primary indicator to classify a WDS function (transmission or distribution networks) is the length-weighted average pipe diameter. The average nodal demand and histogram of total length of each pipe diameter are applied as secondary measures. A new parameter, defined as the branch index (BI), is used to further classify a branched network by estimating the degree of branching within a WDS. The degree of looping and a second level of classification are based on the meshedness coefficient (MC), but only after the system is reduced to eliminate nonessential nodes. BI values are compared with other system-structure metrics in the literature including link density (LD), average node degree (AND), MC, and clustering coefficient (CC). To that end, Pearson correlation coefficients are computed across the set of other metrics for 26 systems. The correlation analysis reveals that several graph-theory system-structure metrics are highly correlated.