Research On Hydraulic Model Establishment Of Urban Underground Water Sypply Network And Leak Detection And Location

The leakage of the water supply network is still relatively serious,which result in the waste of fresh water resources and secondary pollution of water quality are still common,and which brings economic losses to relevant departments and affects the normal water demand of residents.With the development of smart cities,Smart Water relies on computer technology to realize the intelligent monitoring of water supply network,and uses the Supervisory Control and Data Acquisition（SCADA）system of the water supply network to collect real-time monitoring data of various monitoring points in the network to master the operating status of underground water supply network,and which can effectively improve the leakage situation of the current network system.Relying on the smart water system established in a town in the north,this research takes the J district of the town as the research target,constructed the hydraulic model of the water supply network,and established the leakage detection model and the leakage location model respectively based on the historical monitoring data and simulated leakage data of the district,to realize the detection and location of leakage in the network.The specific work was as follows:（1）Imported the preprocessed basic topology data of the J district into the modeling software EPANET,and determined the network roughness coefficient,node demand,water change coefficient and calculation parameters after simplification,and obtained the initial hydraulic calculation model of the network.Then,the genetic algorithm was used to automatically check the roughness coefficient and the node demand.The check results met the accuracy requirements of 80% of the pressure detection point simulation data and the actual monitoring data deviation not exceeding 2 m and 50% deviation not exceeding 1 m.and the J district was successfully constructed.Finally,by setting different emitters coefficient values,the leakage database of J district was obtained.（2）Optimized DBSCAN（Density-Based Spatial Clustering of Applications with Noise）clustering algorithm parameters Eps and Min Pts through multi-objective particle swarm algorithm,established a pressure anomaly identification model based on DBSCAN algorithm,and used simulated leakage data to determine the leakage detection efficiency at each moment.The pressure data after the historical record accident was input into the model,and the test results showed that the pressure anomaly identification model can effectively identify the leakage event.（3）Used the historical flow data of the J district to establish a flow prediction and detection model based on the robust least squares method,and detected whether the real-time flow monitoring value was abnormal while predicting the normal flow of the time series.By adding additional flow to the flow data under normal operating conditions to simulate the leakage of the pipe,leakage events of different leakage sizes occur at 2:00,8:00 and 15:00.Only after 10% leakage occurs at 15:00,the leakage event was not detected,and the rest of the leakage events were detected.The results showed that the leakage detection effect was good.（4）Used Multi-population Genetic Algorithm（MPGA）to optimize the weights and thresholds of the BP neural network,and used the optimized BP neural network to establish the mathematical model of the pressure value of the monitoring point corresponding to the node and the horizontal and vertical coordinates of the leakage node.After the test sample were input,the node coordinate prediction errors were all within 150 m,indicating that the neural network model was more accurate in identifying the leakage position after the leakage.