水库洪水淹没风险快速预测及避险转移路径规划方法研究OA

In the management of reservoir flood inundation risk events,rapid and accurate prediction of downstream flood evolution is crucial for formulating effective emergency response decisions.Conventional mechanistic models are time-consuming and often fail to meet real-time forecasting requirements.To ad-dress this problem,a flood rapid prediction method is proposed integrating the hydrodynamic model with XGBoost.This method utilizes the hydrodynamic model to simulate various inflow scenarios,thereby gen-erating a training dataset.Then,the physical constraints and key features of flood evolution of the dataset is retained to train the XGBoost rapid prediction model.The results show that the rapid prediction model achieves an R2 greater than 0.98 for inundation depth,with a single prediction computation time of 8.35-9.41 s,significantly enhancing efficiency.Based on the prediction results,a hybrid algorithm is introduced to optimize evacuation routes.This algorithm establishes ten precise evacuation routes for four disaster-stricken points at the downstream of the Changyimiao Reservoir,thereby reducing evacuation time and minimizing risk to the personnel.With guaranteed accuracy,this framework for rapid flood pre-diction and evacuation-route optimization enables real-time flood prediction and efficient evacuation deci-sion-making,providing a technical support for flood risk management.