Abstract: The environmental conditions in the Arctic are complex and harsh, and iceberg drift poses a potential threat to the safety of long-term operation point facilities in the Arctic. This study evaluates the iceberg collision risk of Arctic long-term operation point facilities during stationary operations and proposes a Dynamic Bayesian Network (DBN)-based iceberg drift path prediction method. The method combines the iceberg drift equilibrium equation with the momentum conservation method to construct a network structure and realize the dynamic probabilistic prediction of the iceberg drift path. Through real-time environmental data updating, the model outputs the distance of the iceberg relative to the Arctic long-term operation point facilities, providing a quantitative basis for collision risk assessment. The risk scenario analysis based on the prediction results quantifies the collision risk at each time point using the utility function, which is of guiding significance for improving the risk assessment and safety protection capability of Arctic long-term operation point facilities.

Key words: Dynamic Bayesian Network, iceberg path, Arctic long-term operation point facilities, collision risk assessment