The Study Of Mid-long Team Hydrological Forecasting Model Used In The Three Gorges Reservoir Inflow And The Design & Application Of System

Nowadays, the Three Gorges water conservancy project is the unique Dam in the world, it is the core project to manage, develop, utilize and accomplish optimal operation the Yangtze River'water resource. The Three Gorges conservancy project plays a important role in Flood Control and Drought, power generation, irrigation, navigation. It is a significant work to do well the mid-long team hydrological forecasting in the optimal operation of the Three Gorges Reservoir. So far, the Three Gorges Reservoir'long-term hydrological forecasting is still in the exploratory stage and not mature , so it is of great significance to do research about mid-long team hydrological forecasting in the Three Gorges Reservoir.In this paper, Yichang station is set as the representative station and Nearest Neighbor Bootstrapping Regressive model (NNBR), Mean Generating Function (MGF) and Automatic Regressive model (AR) are used as the basic model of the Three Gorges Reservoir inflow long-term forecast. On this basis, proposed a linear regression modified model and a combined forecast model based on Support Vector Machines (SVM) and Entropy, and optimized the forecast model. Besides, the long-term hydrological forecast system of the Three Gorges was constructed and all kinds of forecast models were applied in the Three Gorges Project. The main results are as follows:(1) The establishment of forecasting methods. Based on currently available hydrological information of the Three Gorges and the hydrological characteristics of the region, NNBR, MGF, AR are selected as the basic model of long-term forecasting.(2) Long-term hydrological forecast system development and programming. Based on the model theory, the three models above are programmed by Delphi. On the basis, combined with the hydrological information, the long-term hydrological forecast system of the Three Gorges was built. The system is a object-oriented application software that includes functions such as hydrological information search and management, hydrological elements analyze and long-term hydrological forecast.(3) Application of the forecasting methods. YiChang station is set as the representative station to do long-term forecast research of the Three Gorges Reservoir inflow. The results show that the three selected forecast method have their own advantages and are all suitable for the Three Gorges Reservoir. Forecast passing rate in each model is more than 94% and meets the standard requirements. But the relative error distribution bandwidths of the results are large. The error bandwidths of the three models are 20.28, 28.03 and 15.97. Therefore, each model's forecast result is not absolutely stable.(4) Optimization of forecast model. As the unstable of single model forecast result, a linear regression modified model and a combined model based on SVM and entropy weight were proposed to optimize the forecast methods. The results show each passing rate of optimized model is 100%. Error bandwidths of the modified model are 18.11, 11.93 and 12.24. Error bandwidths of the combined model are 15.71 and 13.89. Compared to single forecast model, the results are all getting better. Therefore, optimization of the basic model is an effective method to improve forecast accuracy.