Spectral Clustering and Multicriteria Decision for Design of District Metered Areas

Abstract

The design of district metered areas (DMAs) has many factors to be considered to reduce the adverse effect of the closure of valves. A three-step methodology for DMA design is proposed based on spectral clustering algorithm and graph theory. The first step is to calculate the nontrivial eigenvectors based on a weighted graph of the water distribution network. The second step is optimal nodes clustering as DMAs by spectral clustering. The third step is to determine the location of meters and valves by a proposed heuristic method. The multicriteria decision method is then used to determine the best DMAs solution. A simple example network is used to demonstrate the method, and a real-world network, the Wolf-Cordera Ranch, is applied to test the feasibility and effectiveness of the proposed method for DMA design in actual systems. Results show that the proposed method can identify different DMA solutions effectively, and better DMA design can be obtained considering different performance criteria.