Research On Sensor Deployment Of Leakage Monitoring In Urban Water Supply Network

Water supply network is an important civil infrastructure to meet the needs of people’s production and living water.Due to aging,corrosion and human impact of pipelines,pipeline leakage accidents occur frequently,resulting in a series of economic and social problems.Therefore,it is very important to detect the leakage of water supply network and locate the leakage positions accurately.At present,the sensor data processing of water supply network is the common means of leakage monitoring.However,too many sensors will lead to high input cost,and the processing of high-dimensional sensor data is also complex.Therefore,how to reasonably arrange a small number of monitoring sensors in the water supply network is the primary task to achieve global leakage monitoring.The main research content of this paper is the method of sensor arrangement for leakage monitoring of urban water supply network.The arrangement of pressure sensor can be regarded as a feature selection problem.In the case of single point leakage,the node pressure in the pipe network is taken as the feature,and the label of the leaking node is taken as the label;in the case of multi-point leakage,the node pressure in the pipe network is taken as the feature,and the leakage of each node(the value of the leaking node is 1,the value of the normal node is-1)is taken as the label.Whether single label feature selection or multi-label feature selection,the node of the selected features are the installation locations of the sensors.Traditional feature selection methods have their inherent problems,which do not consider the lack of leakage location records and multi-point leakage in the actual operation of water supply network.In this paper,the leakage location of monitoring data is studied.By resetting the label set of leakage location,the problem of semi-supervised feature selection is transformed into the problem of full supervision feature selection,and the layout of leakage monitoring sensors in water supply network is completed.In view of the situation of multi-point leakage at the same time in water supply network,this paper proposes a multi-label feature selection method based on mutual information,which not only reduces the dimension of label space,but also realizes the fast and effective monitoring of multi-point leakage.The semi-supervised feature selection method can effectively solve the sensor placement problem in the case of partial missing labels.In this paper,firstly,a scheme of dividing the monitoring area of water supply network is proposed.The pressure sensitivity matrix of water supply network is constructed by using the pressure matrix under normal condition and leakage condition respectively,and the actual measurement error of the sensor is simulated by randomGaussian noise.By analyzing the change law of the pressure at each node,the pressure sensitivity matrix is fuzzy gathered by using Fuzzy C-Means(FCM)to divide the nodes of the same type of pressure data into the same monitoring area.Then,75% of the location labels are deleted randomly.The semi-JMI method is used to select the representative node in each monitoring area,and install the sensor in each representative node to complete the leakage monitoring layout of the water supply network.The experimental results show that the sensor placement strategy adopted in this paper can ensure higher positioning accuracy and smaller positioning deviation even if the proportion of label missing is large.The multi-label feature selection methods can effectively solve the problem of multi-leak monitoring in water supply network.For large-scale water supply networks,the traditional multi-label feature selection methods have large label dimensions,which leads to the high time complexity of the algorithms;at the same time,the direct processing of the original labels will ignore the local correlation of the label space.In view of the above problems,this paper proposes a multi-point leakage monitoring scheme of water supply network based on LSDR-JMI method.Firstly,the pressure sensitivity matrix of pipe network is constructed.Then,k-means clustering method is used to deal with the two-dimensional leak location label,and one-dimensional clustering vector is obtained as pseudo label.By calculating and sorting the contribution of the original label to the pseudo label,the label column with larger contribution is retained as the optimal sub-label space after dimension reduction.The contribution degree of each characteristic column of the pressure sensitivity matrix to the optimal sub-label space is calculated and sorted,and the characteristic column with larger contribution degree is retained.Sensors are installed at the nodes where the characteristic column is located to complete the multi-point leakage monitoring arrangement of the water supply network.Experimental results show that LSDR-JMI method can significantly reduce the time complexity of the algorithm,and accurately locate the leakage location.