Study On DMA Leakage Detection And Location Of Water Distribution System Based On Deep Learning Model

For most countries,the amount of water leakage in the water distribution system is significant,and it is estimated that about 30% of the water leaks in the water distribution system.With the acceleration of the urban construction process and the increase of the service life of the water distribution system,the leakage of the system occurs frequently,which affects the normal production activities and the daily life of the residents.The leakage of the water distribution system is an important problem faced by the world.In addition to the transformation and management of the water supply network,establishing a great leakage detection and location system for the water supply network is important.In this paper,a deep learning model is established to detect and locate leakage on the basis of independent metering partitions,and a research on leakage detection and location of water supply network that supports decision-making is carried out.The main contents and innovations of the paper are as follows:1.An independent metrological partitioning method based on weighted similarity matrix and spectral clustering is introduced.Aiming at the problems of unreasonable partition and heavy workload in the management of independent metered area(DMA)of water supply network,an independent metered area method based on weighted similarity matrix and spectral clustering was proposed.First,according to the connection structure of pipe sections in the water supply network,the concept of natural neighbors of nodes is introduced,and the factors that need to be referred to in the actual partition are converted into weights and brought into the similarity matrix,so as to establish the weighted similarity matrix of the nodes of the pipe network.Finally,the spectral clustering algorithm is used to complete the independent metering partition of the pipe network.The experimental results show that this method can effectively improve the hydraulic performance of the pipe network system.2.Leak detection of water supply network based on deep learning model is applied.In view of the problems of low efficiency and real-time performance of leakage detection methods in the water distribution system,this paper uses the CNN model,LSTM model and CNN-LSTM model to detect leakage in the water supply pipe network.According to different pipe network characteristics and data characteristics Different models can be selected according to different leakage detection requirements,so as to meet various requirements for leakage detection of water supply network.Among them,the CNN-LSTM model is mainly studied.The model performs leakage detection on the basis of independent metering partitions.Its purpose is to detect whether there is leakage in the DMA partition,and to detect and narrow the location range of the leakage nodes in the pipeline network.The case application shows that the deep learning model in this paper can provide a personalized leakage detection method for the pipeline network,so as to realize the effective detection of the leakage of the water supply network.3.A method for leakage detection and location of water distribution system based on CNN-LSTMMSNet-Attention model is introduced.Due to the seasonal characteristics of water pressure and other time series data in the water supply network,it brings challenges to leakage detection and location,resulting in low efficiency and low accuracy of leakage detection and location training.Based on the damage detection results,it focuses on the multiseasonal characteristics of water pressure at nodes in the water supply network.The Long Short-Term Memory Multi-Seasonal Network(LSTMMSNet)is proposed to be used for time series data with multiple seasonal patterns in the water supply network,and an attention mechanism is introduced to distinguish different influencing factors of the leakage of the pipeline network,so as to assist the model to make better changes.Accurate judgment.The case application shows that the CNN-LSTMMSNet-Attention model can accurately capture the multiseasonal characteristics of the data in the water distribution system and find the leaking nodes.And formed the step-by-step research by CNN-LSTM model and the CNN-LSTMMSNet-Attention model for the water supply network leakage detection and location,which improved the accuracy of the leakage detection and location.4.Introduce.a research on leakage detection and location of water distribution system to support decision-making.Aiming at the problem that the leakage detection and positioning system of the current water supply pipe network relies too much on manual labor,this paper conducts a research on the leakage detection and positioning of the water supply pipe network that supports decision-making.First,a database is established to store the relevant data in the water supply pipe network,and then a leakage detection With the positioning of the model library and method library.It establishes the basis for supporting decision-making for the leakage detection and location management of the water supply network,so as to realize the effective detection and location of the leakage of the water supply network.After analyzing the actual leakage detection and positioning requirements of the water supply pipeline network,taking the pipeline network of a coastal city in H area in eastern my country as the application background,using Java Web,Python language and My SQL database and other technologies to build a platform for a coastal city H in eastern my country.The leakage detection and location of the water supply network in the district provides an effective method.Compared with the traditional leakage and location management methods of the water supply network,the application of this platform can effectively manage the leakage in the water supply network,reduce labor costs and improve work efficiency.