Research On Long-term Runoff Forecast And Short-term Optimal Scheduling For Hydropower Stations

Hydropower is the relatively clean, renewable and high-quality energy, and it can be directly converted to a secondary energy, that is electricity. Compared to other energy, hydropower has some advantages, such as low generation costs, rapid unit commitment, strong peaking FM capability, efficient emissions reduction and so on. So both at home and abroad, people attach great importance to the development of hydropower. Recently out country is launching and implementing the "energy saving","west to east" and other policies, and the steady and continuous promotion of hydropower plays an important part in practicing these policies. At present the development and construction of Hydropower shows the characteristics of large-scale, complicated and so on, therefore it puts higher requirements for the optimal allocation of resources, and the scheduling of cascade hydropower stations. At present, the basic situation of Chinese developed hydropower is low efficiency and large waste, so in order to solve the problem, this paper has done a strict and in-depth research taking the hydropower reservoirs (group) as the study object. As is known to all, the long-term runoff forecast and the short-term optimization scheduling has an important position in theory and practice, so research on the two aspects has important implications for perfecting reservoir management theory, improving water use efficiency, and promoting the sustainable development of social, economic and the ecological environment. The main achievements of the whole paper are as follows:(1) The activation function of BP network and its application in runoff prediction model. The traditional BP algorithm has some inherent defects, such as long training time, slow and local convergence and so on. Through analyzing each essential element of the BP network structure in depth, it finds that the activation function factor plays an important role in modeling and model effect. Taking the commonly used activation function in the current BP network model and its improved model as the research object, it designs the mixed level of the whole permutations test program, establishes the appropriate evaluation system, and choses the range analysis method to analyze the sensitivity of the activation function. The results of the examples shows that the different combination of activation function in the layers has important influence on the convergence rate, the convergence precision, and the generalization ability of forecasting model. So a reasonable combination of the hidden layer and output layer activation function can not only save the simple structure and computational efficiency of BP model, but also improves the accuracy of certain predictions. (2) The medium and long term runoff combination forecast model based on the sensitivity analysis of BP activation function. BP network model is often selected as one of the factors in combination forecasting model. Taking the of long-term runoff combination forecast model for study, it introduces the forecast accuracy and stability two evaluation index, and use the objective entropy weight method and subjective G1to determine the relative importance of the two indicators, and then take the variance-covariance method to determine the index weight of each single model, finally gets the combination weights of each single forecasting model. On this basis, the long-term runoff forecasting model has been established based on a sensitivity analysis of the BP network activation function. In the example application, choosing the mixed level of the whole permutation test program and range analysis method to solve the model, and the results have given the based reasonable suggestions on how to choose the activation function combination of BP network. The suggestion saves the characters of BP algorithm, such as simple, logical, and high efficient calculation, simultaneously it improves the prediction accuracy and stability of the forecast model.(3) Self-optimization simulation model of short-term cascaded hydroelectric system dispatching based on the daily load curve. Short-term optimization scheduling of hydropower stations has the great significance in practical production and the implementation scheduling, because it directly connects the long-term optimization scheduling with the plant economic operation. In the actual grid scheduling, hydropower power system has the tasks of FM and peaking, so from the electrical side, the power output progress should be consistent with the system load curve as far as possible; and from the generation side, it should try to improve the efficiency of power generation and peaking. In order to take into account the need of the electrical power requirements of both sides, the optimization simulation model based on the load curve has been proposed. Combining optimization with simulation technology, the automatic identification of feedback has been embedded into the normal simulation model, with the advantages of the self-optimization simulation technology) a four layer identification feedback structure and the optimization structure have been set up to determine the short-term optimal scheduling. And taking the middle reach of the Jinsha River reservoir as an example for simulation calculation, the application shows that the research has a certain degree of practicality and feasibility.(4) Research on short-term scheduling of two-dimensional unsteady flow mathematical model. Hydropower system with multi-objective analysis of flood control, navigation, power generation and other requirements, should formulate a reasonable scheduling scheme to achieve the coordination of the target objective, which is of great significance to make full use of the water resources. In Jinghong-the olive dam reservoir system, the requirements of river navigation between the two dams is V, at the same time, the olive dam reservoir is the last regulating reservoir in Lancang River, and its exit runoff will be restricted by the international rivers flow restriction. From the point view of the two target, a two-dimensional hydrodynamic unsteady flow model is established considering the change of river flow field. Considering the conditions of the joint operation of Xiaowan and Nuozhadu Hydropower, seven typical daily operation scheme have been established for Jinghong-Olive dam. The model selects hybrid unstructured grids to discrete computational domain, uses the approximate Riemann of Roe format solution to calculate the interfacial normal flux, finally it recommends scheduling scheme with practical application value for the joint operation of Jinghong-olive dam reservoir system.