Research On Optimization Operation Models For Cascaded Hydropower Reservoirs Considering Medium Range Forecasting Inflow And Uncertainty

The hydropower energy has becomes one of the obstacle factors to restrict the sustainable development of society and economy. Hydropower energy, as clean energy, has become an important part of energy structure in China. Improving the efficiency of hydropower generation is one of the most significant ways to improve the utilization of water resources. In recent years, with the development of weather forecast technology and computer technology, increasing the forecast horizon is possible with the Quantitative Precipitation Forecasts (QPFs) information. Then, how to improve the efficiency of reservoir hydropower generation according to the existing technology is the most important issue.Thus, this dissertation takes China’s Hun River cascaded hydropower reservoirs as case study, and uses the Quantitative Precipitation Forecasts information of the Global Forecast System (QPFs-GFS). In the dissertation, the inflow forecasting models and the hydropower optimization operation models are studied, respectively. Firstly, the inflow forecasting models in the study are constructed, and medium term forecasting inflows are forecasted based on the QPFs-GFS. Then the hydropower optimization operation models are constructed by combining the QPFs-GFS and inflow forecasting information. The primary achievements of this dissertation are as follows:(1) In hydrologic model, the accuracy of forecasting inflow is affected by the initial values of underlying surface condition, model structure and parameters. Thus, four different types of hydrological models are constructed. Based on the federated filter algorithm, the forecasting inflows of the models are integrated. In this way, the stability and accuracy of the forecasts are enhanced.(2) In the hydropower operation model for cascaded reservoirs system, the forecasting inflows are combined implicitly. Firstly the algorithms of rough set and decision tree are constructed to mine the operation rules. Then the ideal processes of hydropower generation operation are optimized by deterministic optimization algorithm. Based on the ideal processes, the operation policies are mined by the rough set and decision tree, respectively. In the operation policies, the forecasting information of precipitation and inflow are combined as condition attributes.(3) According to the parameterization-simulation-optimization method, the hydropower operation graph is optimized with forecasting information. In the operation graph, the storage and inflow are combined by converting the storage into inflow dimension. In this way, there are two advantages for the operation graph. The first one is reducing the high complexity and computational costs. The second one is the objectives, e.g., efficiency and reliability, are considered in the optimization algorithm by simulation.(4) In the hydropower operation model for cascaded reservoirs system, the forecasting inflows are combined explicitly. Firstly the aggregation-disaggregation method is applied to the cascaded hydropower reservoirs. Based on this method, an equivalent reservoir is established by aggregating the storage and flow of the reservoirs. Based on the equivalent reservoir, the Aggregation-Disaggregation Bayesian Stochastic Dynamic Programming is constructed and the operation policies of equivalent reservoir are optimized. The aggregate output is obtained from the policies according to the storage and forecasting inflow. In real time operation, the aggregate output disaggregated into each reservoir taking, minimization of the spillages and maximizeation of the storage of cascaded reservoirs system as objective function.(5) In order to analyze the performance of hydropower generation operation under uncertainty of forecasting information, the Rolling Horizon Control model is constructed. Based on this model, the variations of efficiency and stability are evaluated with forecast horizon and decision horizon extending under the observed and forecasting inflows, respectively.(6) According to the performance of hydropower generation operation under uncertainty of forecasting information, the forecasting inflow is divided into two parts in the optimization model. The inflows in the first5days are assumed to be accurate, and the inflows in the second5days are assumed to be of high uncertainty. On the basis of the hypothesis, the Two Stage Bayesian Stochastic Dynamic Programming (TS-BSDP) model is presented. The simulation results demonstrate that the performance of TS-BSDP is superior to that of AD-BSDP and other models.Finally, a summary is given and some problems to be further studied are discussed.