Study On Reservoirs Group Operating Simulation Method And System Design Based On Deep Learning Method

In recent decades,with the development of China’s economy,a number of water conservancy projects have been planned and constructed in each major river basin of China.However,while these water conservancy projects undertake tasks such as flood control and power generation,they also inevitably change the hydrological situation of the river basin,so it is necessary that mastering reasonable hydrological laws of river basin,which can affect the direction of the national macro policy profoundly.Therefore,this paper focuses on the hydrological changes of river basin under the influence of water conservancy projects,and makes in-depth research from the aspects of runoff prediction,reservoir operation rules extraction,river basin simulation with deep learning methods.Through these studies,the hydrological changes of the river basin can be found out,which provides a reference for decision makers to further formulate river basin planning.The main contents and innovations of the paper are as follows:(1)In general,it is difficult to predict the extreme value of runoff,to solve the problem,a runoff prediction model based on time series is established in this paper,SA-CEEMDAN-LSTM.The algorithm forecasts runoff according to the idea of "Optimization-Decomposition-Prediction-Synthesis".By decomposing the original runoff time series with mixed modes into sub series with single component,improving the fitting performance of the prediction algorithm.In this paper,the runoff sequence of Panzhihua hydrological station in the middle reaches of Jinsha River and Xiaodeshi hydrological station in Yalong River are taken as prediction objects.The results show that the MAE of the runoff prediction model established in this paper are all within 10%.(2)Regarding the difficulty of real-time sharing of reservoir scheduling plans in the basin,this paper proposes an improved depth neural network method based on adaptive moment estimation parameter optimization algorithm(SA-Adam-DNN)to simulate reservoir operation.By optimizing the parameter updating mode and structure of the depth neural network,the fitting accuracy and robustness of the network are improved.In this paper,the Guanyinyan reservoir in the middle reaches of Jinsha River,Jinpingyiji reservoir and Ertan reservoir in Yalong River are taken as the research objects,the ten day scale dispatching rules of each reservoir were extracted,and each reservoir’s dispatching operation process were simulated.The results show that the reservoir simulation model can accurately describe the operation conditions of different reservoirs.(3)To compensate for the limitations of the current runoff forecast model that difficult to consider the impact of reservoirs,this paper establishes a river basin simulation model and system,which can simulate any watershed according to the "Prediction-Correction-Simulation" process.This simulation model takes into account the impact of hydrological forecasting and hydraulic engineering stress on runoff,and guarantees the continuous of runoff when simulating and predicting.Further,this paper proposed a simulation system framework of the upper reaches of the Yangtze River Basin based on micro service architecture,which realizes the separation of different logic business,front-end business and back-end business.The river basin simulation framework constructed in this paper has been applied to the water resource management decision support system of Jinsha River-Three Gorges cascade hydropower station,and has achieved good results.