The Research Of Urban Water Supply Network Real-time Modeling And Leakage Events Detection Location

A rapid and effective pipeline leakage detection and location method has considerable significance in ensuring the safety of urban water supply and conservation of water resources.Centering on the problem of leakage location of urban water supply network,this paper launches the research on the establishment of DMA real-time model,the prediction of network pressure and the detection of abnormal working conditions,the initial location and accurate location of leakage.The main contents of this paper are as follows:1.The establishment of real-time hydraulic model of urban water supply network.Based on the EPANET-RTX real-time modeling framework,taking network information and SCADA data as the basis in a large DMA YC District of SX,a real-time hydraulic model of water supply network is established,which reduces the average fitting error of the monitoring point from less than 10% of the off-line model to less than 5% and improves the simulation precision.2.Pressure prediction of water supply network based on the LSTM(Long Short-Term Neural Network)model.Aiming at the highly complex nonlinear characteristics of water supply network,a deep learning model LSTM is adopted to predict the status of network.The input of the model includes the status information of the monitoring point pressure and the control information of the water supply pressure and water supply amount at each water inlet.In order to improve the information feature extraction and prediction accuracy,this paper proposes a model of combining the parallel LSTM and the deep neural network(DNN)which is called PLDNN.The test results show that the RMSE of the PLDNN prediction model is only 0.0017,and the MAPE is only 0.45%,which has higher learning performance and accuracy than the traditional prediction method(BP neural network,support vector machine,etc.)and the common LSTM model.Finally,the difference between the predicted value of PLDNN and the measured value is put forward as the basis for judging whether the network occurs abnormal event or not.3.Two kinds of initial leakage location methods.Method 1: according to the abnormal pressure ratio of the monitoring point after the occurrence of the leakage,the boundary point is determined,and the leakage area is defined according to theboundary point.Method 2: firstly,according to the sensitivity analysis and clustering algorithm,the water supply network is divided into a number of "virtual leakage zone",a large number of leakage data are generated through the water supply network model in different leakage zones,the nonlinear mapping relationship between the leakage data and the corresponding leakage partition is established through the depth belief network to achieve the initial location.4.Accurate location of leakage based on real-time hydraulic model.Establish a real-time hydraulic model under the leakage condition,the traditional single target genetic algorithm GA and the multi-objective genetic algorithm NSGA2 are designed,and the hydraulic model is used to check and find the optimal location of the leakage.The result of 5 times leakage localization show that the real-time hydraulic model of leakage is more practical than off-line model,and multi-objective genetic algorithm is better than single objective genetic algorithm.