摘要
通过K均值聚类算法、层次聚类算法、谱聚类算法3种聚类方法,将节点坐标和节点压力作为数据集、轮廓系数作为评价指标,对Anytown和KY32个管网进行了压力分区,并对3种分区方法的压力分区结果在2个模型上进行了比较。通过计算分别得到K均值聚类算法、层次聚类算法、谱聚类算法在Anytown和KY3的最佳轮廓系数分别为0.40336、0.40336、0.39912,0.44833、0.45177、0.45306。通过绘制压力分区结果图,发现层次聚类算法在Anytown管网下与K均值聚类算法一致,但在KY3管网下出现压力分区节点分散的现象。谱聚类算法在Anytown管网存在某个节点需要手动调整问题,在KY3管网同样出现压力分区节点分散的问题。K均值聚类法在2种管网下的压力分区结果较好,各压力分区之间边界明显且压力分区中节点较为紧凑。但3种算法由于无法考虑拓扑结构管道的问题,均需要在压力分区完成后进行调整。
Three clustering methods,namely K-means clustering algorithm,hierarchical clustering algorithm and spectral clustering algorithm,were used to partition the pressure of Anytown and KY3 pipe networks,taking node coordinates and node pressure as datasets and contour coefficient as evaluation index.The pressure partitioning results of the three partitioning methods were compared on the two models.The optimal contour coefficients of K-means clustering algorithm,hierarchical clustering algorithm and spectral clustering algorithm in Anytown and KY3 were respectively 0.40336,0.40336,0.39912,0.44833,0.45177,0.45306.By drawing the result map of pressure partition,it is found that the hierarchical clustering algorithm is consistent with the K-means clustering algorithm under Anytown pipe network,but the phenomenon of pressure partition nodes dispersion occurs under KY3 pipe network.The spectral clustering algorithm has the problem that a node needs to be manually adjusted in Anytown pipe network,and the problem of dispersed nodes in pressure zones also occurs in KY3 pipe network.The K-means clustering method has good results in pressure partitioning under the two pipe networks,with obvious boundary between pressure partitions and relatively compact nodes in pressure partitions.However,because the three algorithms cannot consider the problem of topological structure pipeline,they all need to be adjusted after the pressure partition is completed.
作者
何立新
范一飞
雷晓辉
王琦
HE Li-xin;FAN Yi-fei;LEI Xiao-hui;WANG Qi(School of Water Conservancy and Hydroelectric Power,HUE,Handan 056000,China;Hebei University of Engineering,Hebei Key Laboratory of Intelligent Water Conservancy,Handan 056038,China;School of Civil and Transportation,Guangdong University of Technology,Guangzhou 510006,China)
出处
《海河水利》
2023年第9期73-77,共5页
Haihe Water Resources
关键词
供水管网
压力分区
K均值聚类算法
层次聚类算法
谱聚类算法
轮廓系数
water supply network
pressure zoning
K-means clustering algorithm
hierarchical clustering
spectral clustering
Silhouette Coefficient
