Optimal sensor placement for leak location in water distribution networks based on EGAE clustering algorithm

In the water supply network system, the placement of pressure sensors is crucial to the optimization of the monitoring effect of the system running condition. Most of the existing studies select the placement of sensors according to the hydraulic state of the water distribution network, but do not fully consider and analyze the connectivity between nodes and the topology of the water distribution network. Aiming at the shortcomings of the existing pressure monitoring point layout methods, this paper proposes a pressure sensor layout method based on the Embedding Graph Autoencoder (EGAE) clustering method. Firstly, the EGAE deep clustering algorithm is used to cluster the network nodes based on the topology structure and hydraulic characteristics of the network. According to the clustering results, the network nodes are divided into different regions. Then, by calculating the correlation between nodes in the region, the representative nodes of each region are selected and set as sensor layout nodes. By comparing and analyzing the performance of experiments with other schemes, the results show that the EGAE deep clustering network can effectively integrate the topological structure characteristics and hydraulic characteristics of the water distribution network. The sensor nodes laid out in the scheme proposed in this paper can effectively carry out accurate leakage positioning in subsequent positioning models, and have higher leakage pipe segment identification accuracy and smaller precise positioning errors compared with other schemes.