Pipe Burst Identification In Water Distribution System Based On Real-time Nodal Demand Calibration

Underground burst pipe in water distribution system not only leads to a large amount of waste of water resources,but also may cause secondary pollution of supply water.Timely identification and repair of broken pipes is of great significance for water companies to reduce the gap between production and marketing and improve the quality of service.However,when the pipe burst occurs in remote suburbs or late at night,the traditional manual inspection or customers’ contacts can not timely feedback the pipe burst information,thus delaying the repair time of the pipe burst.Therefore,the real-time online identification of pipe burst in water distribution network has important research significance.In this paper,the burst identification research of urban water distribution network based on real-time calibration of nodal demand is carried out.The specific research contents and achievements include:1)The status quo of China’s water distribution network burst and review of relevant research progress;Firstly,the status quo of pipe burst in China’s water distribution network is briefly introduced and the causes of pipe burst in China’s water distribution network are analyzed.Secondly,the practical significance of real-time online identification of pipe burst in water supply network is introduced,and the research progress of pipe burst identification in water distribution network is reviewed.Finally,through the analysis of the difficulties faced in the current research,the main research content of this paper is drawn out,that is,the real-time online identification of pipe burst in water distribution network is realized through the real-time nodal demand calibration in the hydraulic model of water distribution network.2)The basic principle of hydraulic calculation of water distribution network;As the background knowledge of real-time calibration of nodal demand in water distribution network,the second chapter introduces the basic principle of hydraulic calculation of water distribution network,that is,the construction and solution of hydraulic equation of water distribution network based on mass and energy conservation equation.In addition,the global gradient method for hydraulic calculation of water distribution network is introduced in detail.3)Real-time calibration of nodal demand in water distribution network based on adaptive truncated singular value decomposition method;Since the number of water pressure and flow monitoring points in the actual network is far less than the number of nodes in the pipe network,calibration of nodal demand in the pipe network is an underdetermined problem.In order to solve the underdetermined problem,the current research generally adopts the parameter grouping method,that is,by grouping nodes and assuming that the nodal demand in the same group has the same water consumption pattern,and then reducing the number of unknown quantities,the underdetermined problem is converted to determined or overdetermined for solving.However,the above method has the problem that node grouping is not unique.Using truncated singular value decomposition method to solve the underdetermined problem can effectively avoid the problem that the node grouping is not unique,but it faces the problem of how to determine the appropriate truncated parameter value.In view of this,the third chapter has carried out study of real-time nodal demand calibration based on the adaptive truncated singular value decomposition method,first of all,structure the objective function using the coefficient of variation of nodal demand estimation,and use golden section method to seek the optimal truncation parameter value which gets better tradeoff between data resolution and the stability of the nodal demand estimation.The proposed method is applied to a small pipe network and a real pipe network(called L-Town network),and the results show that the proposed method can achieve a good trade-off between the data resolution and the stability of the solution.4)Pipe burst detection of water distribution network based on the calibration residuals.calibration residual is the difference between the simulated value and the monitored value at the pressure monitoring point after nodal demand calibration.When a pipe burst accident occurs in the pipe network,the calibration residual at the pressure monitoring point near the pipe burst will increase or decrease significantly.According to the relationship between the calibration residuals and pipe burst accidents,the research on pipe burst detection of water distribution network based on the calibration residuals is carried out in Chapter 4.In order to reduce the influence of random fluctuation of nodal demand and monitoring error on the calibration residuals,moving average filtering was used to de-noise the calibration residuals to obtain the filtered residuals.In order to realize the online detection of burst events,the residual threshold is set according to the historical calibration residual data.When the calibration residual of a pressure monitoring point exceeds the set threshold,the monitoring point will trigger a pipe burst alarm.The proposed method was applied to the pipe burst detection of L-Town network.The results show that for 19 pipe bursts,10 of them were detected by the proposed method,and the detection rate of pipe burst was 52.63%.In particular,for the abrust pipe burst,the proposed method can detect it quickly and the detection time is only about 10 minutes later than the beginning time of abrust pipe burst.It can be seen that the burst detection method proposed is feasible.5)Pipe burst location of water distribution network based on improved cosine similarity method.The traditional pipe network burst location method based on cosine similarity determines the burst location by comparing the cosine similarity between the pressure residual vector and each column(corresponding to each node)of the pressure sensitivity matrix.However,this method assumes that the leakage is located on the node,so the location result is not accurate enough.In this paper,the following improvements are made:(1)the sensitivity matrix of node pressure to nodal demand in the original method is replaced by The sensitivity matrix of node pressure to pipeline flow,which makes the method more applicable to the actual situation(the leakage is located on the pipeline);(2)In order to reduce the interference of other pipe bursts on the location of the pipe bursts,the pressure residual vector is composed of the calibration residual of the monitoring points that trigger the pipe bursts alarm.The method was used to locate the detected burst pipes in L-Town pipe network,and the feasibility and effectiveness of the method were verified.All of the algorithms proposed in this article are implemented programmatically in Matlab environment,and applied to the battle of leakage detetion and isolation method initiated by CCWI/WDSA 2020 conference.research results are employed by CCWI/WDSA 2020 conference and are published on the website(https://doi.org/10.5281/zenodo.3925507).And these results will be submitted to the Journal of water resources planning and management(SCI journals).