2021/04/28 by Gergely Hajgató, Hajgató, Gergely, Bálint Gyires-Tóth +3 · 1 citation
Computer Science · Engineering · Materials Science · #Anomaly Detection Techniques and Applications #FOS: Computer and information sciences #High voltage insulation and dielectric phenomena #Machine Learning (cs.LG) #Water Systems and Optimization
paper · pdf · doi:10.48550/arxiv.2104.13619
openalex publication_date 2021/04/28 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
Knowing the pressure at all times in each node of a water distribution system\n(WDS) facilitates safe and efficient operation. Yet, complete measurement data\ncannot be collected due to the limited number of instruments in a real-life\nWDS. The data-driven methodology of reconstructing all the nodal pressures by\nobserving only a limited number of nodes is presented in the paper. The\nreconstruction method is based on K-localized spectral graph filters, wherewith\ngraph convolution on water networks is possible. The effect of the number of\nlayers, layer depth and the degree of the Chebyshev-polynomial applied in the\nkernel is discussed taking into account the peculiarities of the application.\nIn addition, a weighting method is shown, wherewith information on friction\nloss can be embed into the spectral graph filters through the adjacency matrix.\nThe performance of the proposed model is presented on 3 WDSs at different\nnumber of nodes observed compared to the total number of nodes. The weighted\nconnections prove no benefit over the binary connections, but the proposed\nmodel reconstructs the nodal pressure with at most 5% relative error on average\nat an observation ratio of 5% at least. The results are achieved with shallow\ngraph neural networks by following the considerations discussed in the paper.\n