Using complex network theory for missing data reconstruction in water distribution networks

Abstract

Dealing with missing data is a critical challenge in analyzing water distribution networks (WDNs). These unknown data are often associated with the physical characteristics of pipes, like diameter, which are the foundations of several modeling endeavors. This study develops a fully automated model to reconstruct missing diameter information in WDNs within the complex network theory (CNT) framework. The newly developed model employs customized graph measures to systematically retrieve missing diameters based on the topological (e.g., connectivity) and hydraulic (e.g., flow) features of edges with available information. The proposed model is validated by testing it on three real-world WDNs located in Iran and Austria, where data gaps are created by randomly and progressively eliminating the pipe diameter information. Subsequently, the model's accuracy is assessed by comparing the diameter, pressure, and hydraulic-based resilience of the reconstructed WDNs with those of the complete dataset. The results indicate that the proposed reconstruction model can successfully retrieve missing information up to 90% of the data gap. In addition, the model is highly computationally efficient and can quickly reconstruct thousands of missing diameters. Our novel approach can benefit water utilities that frequently encounter incomplete data in their models and require missing data retrieval from large-scale WDNs.