Simulation And Prediction Methods Of Water Level In The Middle Route Of South-to-North Water Transfer Project And Application

The middle route of south-to-north water transfer project(MRP)is a national strategic facility to optimize the allocation of water resources,relieve water shortages in Beijing,Tianjin,Hebei and Henan provinces,promote social and economic development and ensure ecological and environmental benefits.The MRP adopts the operating mode of constant water level of immediately upstream of the gate.Due to the operation characteristics of strict control of water level and difficulty of hydraulic control,it is difficult to make smooth operation and real-time decision.Therefore,accurately predicting the water level of immediately upstream of the gate is the first premise to control the overall water level of the channel section and realize the real-time dispatching decision.Aiming at the requirements of real-time scheduling and decision response of MRP,this thesis explores the applicability of the hydrodynamics and data-driven simulation and prediction methods for water level of immediately upstream of the gate in MRP under the existing data conditions.Combined with daily scheduling and management requirements and integrated models,the "Early warning of water level system of MRP"is designed to support the operation management and real-time decision-making of the MRP.The main contents and results are as follows:(1)The water level simulation method based on hydrodynamic simulation.By generalizing boundaries of check gates,inverted siphons,escape gates,the one-dimensional hydrodynamic simulation model of MRP based on the saint-venant equations is built.The water levels of immediately upstream of the gates in Jingshi section of MRP are simulated under the condition of the existing data,and the applicability of the model are analyzed.(2)The simulation and prediction method of water level in MRP based on deep learning.The TensorFlow deep learning framework is adopted to construct three real-time prediction models with different structures of water level of immediately upstream of the gates,and Adam gradient optimization algorithm is adopted to accelerate model's convergence.The water levels of immediately upstream of the gates in Jingshi section are simulated and predicted,which realizes the real-time forecast of water level.By selecting the optimal model of water level prediction and analyzing the sensitivity of model parameters,the influence of model hyper-parameters and forecast periods on prediction accuracy and calculation efficiency is studied.(3)The "Early warning of water level system of MRP"and application.Based on the operational process and practical requirements of the daily operation on MRP,a hierarchical warning method of the water level of immediately upstream of the gate and real-time scheduling rules are proposed.Based on the comparative results of the adaptability of hydrodynamic simulation model and deep learning water level prediction models in MRP,the real-time water level prediction method based on deep learning is used as the early warning method of the water level of immediately upstream of the gate in MRP,and The "Early warning of water level system of MRP" is designed to support the daily operation scheduling and real-time decision-making.