基于BP神经网络的管网漏失定位研究—以H市为例  

A study of a BP neural network-based leak location model for pipe networks—taking city H as an example

作　　者：王书盛 吴永强[1,2] WANG Shu-sheng;WU Yong-qiang(Department of Municipal and Environmental Engineering,Hebei University of Architecture,Zhangjiakou,075000;Hebei Key Laboratory of Water Quality Engineering and Comprehensive Utilizationof Water Resources,Zhangjiakou,075000)

机构地区：[1]河北建筑工程学院市政与环境工程系,河北张家口075000 [2]河北省水质工程与水资源综合利用重点实验室,河北张家口075000

出　　处：《河北建筑工程学院学报》2022年第1期120-126,共7页Journal of Hebei Institute of Architecture and Civil Engineering

摘　　要：利用EPANET建模软件及H市某区智慧供水平台收集数据,分析研究城市供水管网泄漏事故,将泄漏发生时节点压力变化与泄漏点位置之间建立映射关系,通过水力模型与BP神经网络系统动态分析。结果表明:当漏损面积较小时,漏损定位的效果最好,预测点与漏损点的最近距离为19.71 m,最远距离为192.29 m,平均偏差达78.1 m,模型具有较高的精确性,当泄漏量扩大时,模型的定位精度有所下降,但偏差值仍能维持在300~400 m左右。Based on the data collected by EPANET modeling software and the intelligent water supply platform in a district of H City,the leakage accident of urban water supply network is analyzed and studied.The mapping relationship between the change of node pressure and the location of leakage point is established,and the dynamic analysis is carried out through hydraulic model and BP neural network system.The results show that the effect of leakage location is the best when the leakage area is small;The shortest distance between the prediction point and the leakage point is 19.71 m,the longest distance is 192.29 m,and the average deviation is 78.1 m.The model has high accuracy.When the leakage is expanded,the positioning accuracy of the model decreases,but the deviation can still be maintained at about 300~400 m.

关 键 词：EPANET 水力模型 BP神经网络 泄漏点 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TU991.33[自动化与计算机技术—控制科学与工程]