基于深度学习框架的时空联合供水管网漏损检测研究  

Combined spatial-temporal leakage detection within water distribution system based on deep learning framework

作　　者：蒋白懿[1] 牟天蔚 李维轲 王康 肖敏 王鑫 JIANG Baiyi;MU Tianwei;LI Weike;WANG Kang;XIAO Min;WANG Xin(School of Municipal and Environmental Engineering,Shenyang Jianzhu University,Shenyang 110168,China;School of Architecture and Civil Engineering,Shenyang University of Technology,Shenyang 110870,China;North China Municipal Engineering Design&Research Institute Co.,Ltd.,Tianjin 300074,China;Key Lab of Eco-restoration of Regional Contaminated Environment,Ministry of Education,Shenyang University,Shenyang 110003,China)

机构地区：[1]沈阳建筑大学市政与环境工程学院,沈阳1110168 [2]沈阳工业大学建筑与土木工程学院,沈阳110870 [3]中国市政工程华北设计研究总院有限公司,天津300074 [4]沈阳大学环境学院区域污染环境生态修复教育部重点实验室,沈阳110003

出　　处：《给水排水》2024年第6期152-158,共7页Water & Wastewater Engineering

基　　金：沈阳市科学技术计划(22-322-3-14)。

摘　　要：以深度学习框架为基础,提出了一种时空联合供水管网漏损检测模型。该模型首先运用Node2Vec算法求解不同时间段内节点特征;其次,通过模糊C-均值聚类法,利用管网模型节点特征进行分区。最后,以不同时间段的压力敏感度作为输入,漏损位置的分区号作为标签,通过深度信念神经网络进行训练,并通过训练后的模型对管网漏损位置进行检测。在实例分析中,以A市实际供水管网拓扑结构进行验证,利用MATLAB-Open Water Analytics toolbox联合编程建模,结果表明,各个时间段的检测效果均较优,正确率均达到为80%以上。因此,该模型能够有效地检测管网漏损。A spatial-temporal leakage detection model was presented based on deep learning framework.Firstly,nodes in water distribution system were divided into different clusters in terms of the topological structure via Fuzzy c-mean(FCM)algorithm.Then,the embeddings of SCADA nodes were solved by Node2Vec algorithm.Finally,the deep belief network(DBN)was employed to train a model that leakage condition was regarded as label while embedding was input.The leakage was detected by the trained model.In the case study,MATLAB and Open Water Analytics toolbox were applied to code the spatial-temporal leakage detection model.The results showed that accuracies were all up to 8o%.Therefore,the model had a great effect on the leakage detection.

关 键 词：Node2Vec 深度学习 漏损定位 随机游走 图嵌入 

分 类 号：TU990[建筑科学—市政工程]