Abstract: In order to develop a decision-making system for water supply network dispatching, the pressure variation of every node and the influence between nodes need to be analyzed which takes a huge workload. Therefore, the group-average method of system cluster analysis is adopted to cluster nodes. One node is chosen from every cluster according to deviation from average. Besides, aiming at disposing the missing values and the inconformity of nodes’ acquisition moments during data collection, a data-preprocessing method is promoted. Test results show that the selected nodes excellently represent their clusters and cover the whole geographical area, which satisfy engineering demand.

Key words: system clustering, group-average method, water supply network, pressure node, data preprocessing

摘要： 在开发供水管网调度决策系统时，需对供水管网数百个节点的压力变化及相互间关系进行分析，工作量较大。针对此问题，采用系统聚类中的类平均法将节点聚类，并利用离均差从各类中筛选出代表点。提出针对数据采集时的缺失值及节点间采集时刻不一致问题的数据预处理方法。经实例检验，得出的节点具有良好的代表性及全面性，可满足工程需求。

关键词: 系统聚类, 类平均法, 供水管网, 压力节点, 数据预处理

CLC Number: 

• TP274