A method for water supply network DMA partitioning planning based on improved spectral clustering

Abstract

Recently, the scale and complexity of water distribution networks (WDNs) have been increasing with the acceleration of urbanization process. It has become a hot research focus to use district metering area (DMA) for more efficient management and control of WDNs. This article proposes a multistage DMA planning method based on improved weighted spectral clustering and genetic algorithm, aiming to address issues such as high investment cost and large differences in the water network demand distribution. First, the actual case pipe network is transformed into an undirected weighted graph based on graph theory, and a similarity matrix is formed by combining the physical properties and hydraulic characteristics of the water network. Then, based on the similarity matrix, the weighted spectral clustering algorithm is used to preliminarily divide the WDN, and the performance of the water supply pipe network formed with different division quantities and different weighting schemes is discussed. Finally, the genetic algorithm is used to optimize the arrangement of valves and flow meters on boundary pipes to generate the final configuration of DMA. The results show that the proposed method has a significant improvement in pipe network topology, hydraulic performance index, and economy compared with the traditional DMA method. HIGHLIGHTS The similarity matrix of multiple data fusion of the water supply network is established to enrich the hydraulic information of the undirected weighted graph.; The weighted spectral clustering algorithm based on k-medoids significantly improves the initial district metering area partition performance.; The equipment layout scheme obtained by using the improved genetic algorithm is the most economical.;

Keywords

• district metering area
• genetic algorithm
• undirected weighted graph
• water distribution networks
• weighted spectral clustering algorithm