基于BP神经网络的DMA漏损定位仿真实验设计

Simulation Experiment Design of DMA Leakage Localization Based on BP Neural Network

独立计量区(District metered area,DMA)技术和漏损控制是供水企业降低供水管网运营成本的重要手段。DMA分区优化技术在一定程度上精简了传感器配置数量,DMA极端简单配置传感器数量条件下的漏损定位方法研究还比较鲜见。同时,现有研究成果对漏损定位方法干扰因素问题的分析还不多见。因此,基于仿真实验对漏损定位的上述两个问题进行研究。首先,介绍供水管网漏损的概念以及DMA技术在供水管网中的应用。然后,基于EPANET-Matlab-Toolkit-2.1.1包含的Net1.inp文件,在EPANET仿真软件平台上采用单一入口节点配置水压和水流传感器模拟有限资源条件下的独立计量区。接着,运行EPANET仿真软件分别模拟理想条件情形和某供水节点需水量临时大增情形独立计量区24小时供水情况,获取两个入口节点数据集。最后,采用MATLAB平台编写BP神经网络预测模型程序,并采用上述两个数据集分别进行训练测试。结果显示理想条件下BP神经网络预测模型在有限资源条件下的DMA漏损定位精度很高,供水节点需水量临时大增情形将增大模型预测误差。

District metered area(DMA)technology and leakage control are important means for water supply enterprises to reduce the operating costs of water supply networks.DMA partition optimization technology has to some extent simplified the number of sensor configurations,but research on leak localization methods under extremely simple sensor configuration conditions in DMA is still relatively rare.At the same time,the existing research results are rare in the analysis of Confounding of leakage location methods.Therefore,based on simulation experiments,the above two issues of leakage localization are studied.Firstly,introduce the concept of water supply network leakage and the application of DMA technology in water supply networks.Then,based on the Net1.inp file included in EPANET Matlab Toolkit 2.1.1,a single inlet node was used to configure water pressure and flow sensors on the EPANET simulation software platform to simulate independent metering zones under limited resource conditions.Next,run EPANET simulation software to simulate the 24-hour water supply situation in the independent metering area under ideal conditions and a temporary increase in water demand at a certain water supply node,and obtain two datasets for the inlet nodes.Finally,a BP neural network prediction model program was developed using the MATLAB platform,and the two datasets were trained and tested separately.The results show that under ideal conditions,the BP neural network prediction model has high accuracy in DMA leakage localization under limited resource conditions,and the temporary increase in water demand at water supply nodes will increase the model prediction error.