The risk assessment model for water diversion projects based on a fuzzy Bayesian network

The water transfer project spans a long distance, passing through various regions, including areas prone to heavy rainfall and floods, geological disaster zones, deep excavation and fill areas, and densely populated regions. It faces multiple risks including sudden pollution, flood overflow, and channel instability. Quickly and accurately identifying and assessing these risks is key to reducing potential losses. However, due to the limitations of available data, as well as its ambiguity, uncertainty, and complexity, existing assessment methods struggle to comprehensively and effectively identify the complex risks of water transfer projects. Therefore, this paper innovatively proposes a Bayesian network model based on fuzzy theory and hedging theory. It refines the risk level quantification through a fuzzy comprehensive evaluation membership function and calculates the likelihood function using hedging theory. The Bayesian network model is structured with risk levels as core nodes, disaster-causing factors as intermediate nodes, and risk accidents as terminal nodes. Finally, the defuzzification is carried out using the centroid method to accurately determine risk levels. The proposed method was applied to simulate typical risk characteristics in the South-to-North Water Diversion Project. Through three representative scenarios, the model’s applicability and risk identification capability under multi-source input conditions were systematically validated. Chi-square tests were performed to assess the statistical significance of each risk factor’s influence. The results demonstrate that the model exhibits high consistency and stability across different scenarios, providing reliable decision support for risk identification and the development of early warning strategies in water diversion projects.