基于谱聚类的供水管网DMA分区方法研究  

Research on DMA partitioning method for water supply network based on spectral clustering

作　　者：董建武 余豪 张浩东 孟雨婷 黄杰俊 DONG Jianwu;YU Hao;ZHANG Haodong;MENG Yuting;HUANG Jiejun(PowerChina Zhongnan Engineering Corporation Limited,Changsha 410014,China)

机构地区：[1]中国电建集团中南勘测设计研究院有限公司,长沙410014

出　　处：《给水排水》2024年第S1期329-333,共5页Water & Wastewater Engineering

摘　　要：为进一步提高供水管网分区合理性,解决因人工经验进行分区后,管网水力性能不明的问题,提出了基于谱聚类、改进的节点自然邻算法和遗传算法的DMA分区方法。该方法对管段权重矩阵进行无量纲化,改进节点自然邻算法,进而利用谱聚类算法完成DMA分区,基于遗传算法优化水表、阀门位置,最终确定DMA分区方案。选择两个不同规模的供水管网对该方法进行了合理性验证。结果表明:权重矩阵无量纲化和节点自然邻算法的改进扩展了原算法的适用性,该方法拟定的DMA分区方案,水力性能未出现明显下降,具备可行性。To further improve the rationality of water supply network partitioning and solve the problem of unclear hydraulic performance of the network after manual partitioning,a DMA partitioning method based on spectral clustering,improved node natural neighbor algorithm,and genetic algorithm is proposed.This method performs dimensionless transformation on the weight matrix of the pipeline section,improves the node natural neighbor algorithm,and then uses spectral clustering algorithm to complete DMA partitioning.Based on genetic algorithm optimization of water meter and valve positions,the DMA partitioning scheme is ultimately determined.The rationality of this method was validated by selecting two water supply networks of different scales.The validation results showed that the non dimensionalization of the weight matrix and the improvement of the node natural neighbor algorithm extended the applicability of the original algorithm;The DMA partitioning scheme proposed by this method did not show a significant decrease in hydraulic performance.

关 键 词：DMA 谱聚类 改进的节点自然邻 遗传算法 

分 类 号：TU990[建筑科学—市政工程]