Research On Leakage Detection And Location District Metering Areas In Water Distribution Systems Based On Spatial And Temporal Features

In recent years,with the long-term service of pipeline network,the leakage problem of urban water distribution networks has become increasingly prominent.Controlling and reducing leakage rate has become an urgent task for various government departments and water supply companies.In order to alleviate the economic loss and social harm caused by leakage problem,it is of great importance to explore practical and effective leakage detection methods in water distribution networks.Existing leakage detection methods are susceptible to the influence of the water consumption patterns in water distribution networks,and lead to a high false alarm rate,besides,it is difficult to effectively detect and locate unsupported leakage,which has a small amount of leakage flow.In response to the above problems,this thesis studies leakage detection and location methods of district metering areas in water distribution systems on the basis of spatial and temporal features.The main content and innovations of the thesis are as follows:(1)To alleviate the influence of water consumption pattern to flow based leakage detection methods and reduce the corresponding false alarm rare,a burst detection method based on Gaussian process regression is proposed.First,the fluctuation pattern of flow in district metering area is analyzed,and the random fluctuation in flow caused by water consumption patterns is explained.Based on water balance in DMA,this thesis uses principal component analysis to extract the main fluctuation patterns of water consumption in DMA,then adopts the gaussian process regression method to predict the inlet flow,which not only provide predicted value,but also provides the corresponding probability distribution to model the random fluctuations of the flow,and finally the burst events are detected based on the Pauta criterion.Experiment results show that this method can effectively reduce the false alarm rate caused by random fluctuations in flow,and has a better performance in regards of burst events detection.(2)To improve the performance of unreported leakage detection due to its low leakage flow,the thesis proposed an unreported leakage detection in district metering area based on temporal and spatial features.First,from the perspective of hydraulics model of water distribution networks,the feature that abnormal pressure fluctuations at multiple monitoring stations caused by leakage are analyzed,based on which,a spatial temporal feature extraction method is proposed,the feature is predicted using convolutional long and short-term memory,finally,the unreported leakage is discriminated by support vector machine.Adopting C-Town benchmark as study case,the results prove that the method in this paper has better robustness and practicability to unreported leakage and has a lower leakage detection limit compared to traditional hydraulic model and principal component analysis method(3)Due to the limited monitoring stations arrangement,it’s difficult to locate the leakage in DMA,to solve this problem,a leakage location method based on feature similarity is proposed.Based on the idea of clustering,the nodes in water distribution networks are aggregated into leakage identification areas in view of pressure pattern,for each identification area,prediction model of spatial temporal features is trained.By evaluating the similarity of predicted and actual feature using cosine distance,the identification area can be determined based on the criterion that lower the similarity is,greater the possibility of leakage can be.The results of study case show that the method in this thesis can achieve the identification of the leakage area effectively after the appropriate selection of the divided areas number.(4)Based on the analysis of the actual leakage monitoring demand of water companies,the leakage monitoring platform of urban water distribution network was designed and developed,the platform framework was built using Java Web technology and MySQL database technology.The designs of database module,algorithm module,backstage module and user interaction interfaces are explained.Finally,adopting a DMA in in South China as test case,the leakage monitoring platform was verified by using fire hydrants to release water as leakage simulation.