Reservoir Operation Mode By Coupling Mid-and Long-Term Runoff Forecast Information

With the development of social economy, the contradiction between supply and demand of water resources is more and more prominent. As the main engineering means of regulation, reservoir plays a more and more important role in the optimal allocation of water resources. In the reservoir operation, due to the objective factors like uncertainty of inflow, the limitation of reservoir regulation and subjective decisions of decision makers, it is easy to cause the waste of water resources. With the development of meteorology and hydrology forecast, mid-and long-term runoff forecast can provide important reference for decision makers, and improve the utilization ratio of water resources. However, the forecast exsits some uncertainty, which may increase the operation risk. Therefore, it is necessary to improve the availability of mid-and long-term runoff forecast in order to improve the reliability of the operation model input. Then the reservoir operation mode should be studied to analyse the risk brought from the uncertainty of forecast. Based on the objectives about water supply of inter-basin water diversion and power generation of hydropower stations, the main content and research results are as follows:(1) The forecast factors selection method in mid-and long-term inflow forecast is discussed by using mutual information (MI) and back-propagation neural network (BPN) model in order to improve the ability of forecast model. The study is applied to Biliuhe reservoir. First, MI and correlation coefficient are used to select input factors. Then, using maximum MI of training samples and texting samples as objective functions to select forecast factors with BPN, and identify the compound correlation between runoff and its impact factors. The objective, mean square error (MSE) is also compared. Finally, the forecast factors are determined, and the validation samples are calculated. The results show the methods of factors selection combining MI with BPN model may provide a good reference for mid-and long-term runoff forecasting.(2) The method of numerical rainfall forecast with different forecast horizon weights is proposed in order to improve availability of ten-day runoff forecast. Biliuhe and Huanren are selected as cases study. Numerical rainfall forecast of Global forecast System (Global Forecasting System, GFS) is used in the study. First, GFS and the actual rainfall data of two areas are sorted. Based on genetic algorithm, weighted rainfall prediction value is calculated. Then, the availability before and after weighting are compared. Finally, a ten-day runoff forecast model is used to illustrate the effectiveness of the weighted GFS method. Results show that the weighted method can improve the availability of numerical rainfall forecast, and improve the availability of ten-day runoff forecast.(3) Based on the bayesian stochastic dynamic programming model, optimal operation mode of inter-basin water diversion by coupling the long-term runoff forecast and its uncertainty is established. With considering forecast uncertainty, the runoff description model is improved based on the forecast information of the wet season. The bayesian stochastic dynamic programming model with considering long-term forecast (BSDP-LTF) is established. Other stochastic dynamic programming models are compared with BSDP-LTF. The results show the runoff description is improved by BSDP-LTF and the effectiveness of the inter-basin water diversion is increased with the present forecast accuracy and risk.(4) Based on parameters-simulation-optimization model (PSO), optimal operation mode of inter-basin water diversion by coupling the mid-term and long-term runoff forecast is established. First, the operation model of inter-basin water diversion is introduced, forecast operation mode set are proposed. The methods of employing long-term forecast to modify top and bottom limitation of the water division rules, and employing "storage plus mid-term ronoff forecast" are proposed. Then, the different forecast operation methods are compared by simulation results. The rationality of coupling the mid-term and long-term runoff forecast information is analyzed. Finally, it is assume that long-term forecast uncertainty is changed, to analyse the operation risk. Results showed that when long-term forecast accuracy achieves a certain level, the coupling mid-term and long-term runoff forecast information operation method can provide reference for the decision.(5) The hydrogen energy storage for cascade hydropower stations with forecast operation is proposed according to the operation risk problem coupling forecast information. Hunjiang cascade hydropower stations is taken as a case, based on the forecast optimization operation rule curves coupling ten-day runoff forecast, hydrogen energy storage for cascade hydropower stations is proposed as a method of reducing the forecast error effect and processing error in risk-decision. By establishing hydrogen production operation rules, the electricity generation reliability of cascade hydropower stations and the utilization rate of hydrogen production equipment are analyzed under different scales of hydrogen production. The results can be used as a reference to choose a reasonable scale of hydrogen production for hydropower stations. The method also adapts to the demand for hydrogen energy in the future and makes full use of water resources with far-sighted.