Multi-level leakage risk management integrated framework for urban water distribution network

Abstract

Global WDN are universally challenged by water leakage and aging, necessitating the adoption of scientific O&M strategies to mitigate risks and ensure reliable operation. Existing research has focused solely on a single dimension and ignored the hierarchical characteristics of WDNs. Traditional models have limitations regarding adaptability to time-series and unbalanced data processing. This study integrates pipe structure characteristics and risk management theory to propose a multi-level leakage risk management framework, dividing the WDN into two levels: main pipes and DMAs. For predicting main pipe failures, the study compares survival analysis, machine learning, and their integrated models, innovatively adopting the integrated model XGBSE to optimize temporal risk modeling. The GADW-RTOPSIS dynamic decision-making method is proposed for DMA risk assessment, dynamically adjusting subjective and objective weights to adapt to data uncertainty and enhance assessment reliability. Tests based on Dongguan City’s WDN show that the XGBSE achieves a C-index of 0.9583 in main pipe failure prediction tasks, significantly outperforming traditional models. The GADW-RTOPSIS demonstrates powerful high-risk sorting capabilities and economic benefits 3.34 times greater than the overall average when guiding DMA renovations. The research findings provide a refined risk-driven decision-making paradigm for urban WDN, supporting the efficient allocation of pipe O&M resources.