基于数据与机理驱动的密云水库洪水预报技术研究及应用OA

As an important flood control project surface drinking water source and strategic reserve base for water resources in the Beijing,the safe operation of the Miyun Reservoir is crucial for the safety of flood control and water resources.In recent years,under the influence of the"natural artificial"binary factor,the production and convergence characteristics of the watershed have undergone significant changes.High water level operation has put forward higher requirements for the accuracy of reservoir flood forecasting and the original flood forecasting model system can no longer meet the requirements.On the basis of the original flood forecasting model system,this article takes the upstream basin of the Miyun Reservoir as the research object,and systematically studies the hydrological simulation technology of high-intensity human activities efficient,parameter calibration technology of hydrological models,data-driven flood forecasting technology and multi model ensemble forecasting technology based on Bayesian averaging.The results show that the mechanism driven model has higher accuracy in flood peak prediction,but it shows a low valley and a lag in peak time.The peak time prediction of data-driven models is more accurate and the overall accuracy of flood peak prediction is not as good as that of mechanism driven models.The Bayesian average integrated with two types of models is close to the actual process and the prediction accuracy is greatly improved.