基于用户水质投诉信息的供水管网污染源的追踪定位  

Location of Contamination Source in Water Distribution Networks Based on Water Quality Complaints

作　　者：信昆仑[1] 刘龙[1] 陶涛[1] 项宁银[1] 

机构地区：[1]同济大学环境科学与工程学院,上海200092

出　　处：《天津大学学报（自然科学与工程技术版）》2014年第4期336-342,共7页Journal of Tianjin University：Science and Technology

基　　金：国家自然科学基金资助项目(50908165)

摘　　要：用户对水质异常情况的投诉是反映供水管网整体水质状况的重要信息,但由于投诉信息具有滞后性、非量化等特征,目前尚未有基于这些信息的管网污染源识别方法.基于模式识别理论,提出了一种根据用户投诉信息追踪定位污染源的方法.该方法首先根据投诉节点的位置信息确定污染源候选节点集合,通过水质模拟确定候选节点发生污染后的用户水质投诉顺序,并以此为基础构建模式识别神经网络.考虑到用户投诉信息的特点,采用有噪声的样本对神经网络进行训练,并对不同类型噪声的样本进行测试.结果表明,训练后的神经网络可以有效地识别用户投诉信息模式,进而确定污染物注入位置.Consumer complaints on water quality anomalies are important information for water companies to understand the service status of water distribution network. However,due to the uncertainties of complaint information about the time and concentration of contaminants arriving at complaint nodes,there are few source identification methods reported based on complaints yet. In this paper,a pattern recognition method is presented to identify the location of pollution source using the consumer complaint data. With the method,a group of candidate nodes are first figured out,which are possible intrusion points. Through a water quality simulator,the time and position patterns of complaints are calculated and coded as the inputs of an artificial neural network（ANN）. Considering the uncertainties of complaint information,the ANN was trained by samples with noises and then tested by different kinds of sample data. It is concluded that the trained neural network is able to recognize the complaint pattern and identify the pollution source location effectively.

关 键 词：供水管网 污染追踪定位 模式识别 神经网络 

分 类 号：TU991[建筑科学—市政工程]