Research on Leak Location Method of Water Supply Network based on Deep Neural Network Model

Abstract

The water supply network is one of the important infrastructure in urban construction. It has strong theoretical and practical significance to realize the real-time monitoring and leak location of the water supply network. In this paper, based on the similarity of water supply network node pressure, fuzzy C-means clustering algorithm is used to realize the selection of finite monitoring points. On this basis, a depth neural network model is constructed according to the pressure changes of the monitoring points before and after the leakage of the water supply network, so as to locate the leakage points. In the experimental part, hydraulics simulation was conducted by using EPANETH pipe network adjustment software according to the layout structure of water supply network, and the pressure of all nodes was obtained. A deep neural network model was established by Keras in Tensorflow framework. After model training and testing, the training error was controlled within the effective range of 5 %. Finally, the model is applied to the actual leakage problem of underground water supply network in Langxi County of Xuancheng City, and the accurate location of the leakage point is realized. The experiment proves the feasibility and accuracy of the method proposed in this paper.