Study On Multi-Objective Optimal Operation Of The Reservoir On Sediment-Laden River

The problem of the multi-objective optimal operation of the reservoir onsediment-laden river is an integrated subject of the flood forecasting, the reservoirsedimentation calculation, and the optimal operation. In this dissertation, thecharacteristics and chaos of the monthly runoff and sediment concentration of theYellow River have been studied. The BP neural network model is improved by thegenetic algorithm (GA), which is used in fast prediction of the height of Tongguanand the reservoir sediment scour and deposition. By analyzing the characteristics ofstream flow and the sediment concentration in a flood process, their pattern can berecognized and simulated. The random differential equation model of the sedimentconcentration has been set up, and then the probability density of the sedimentconcentration and the expected values of the reservoir sediment scour and depositioncan be obtained by resolving this equation. The iteration fuzzy pattern recognitionmodel and GA have been used to operate the reservoir on sediment-laden river. Theprincipal contents and achievement of this dissertation are summarized as follows:1. The improved BP model, which is combined with GA and BP neural network,has been employed to overcome the disadvantages of each other. The initialthreshold value and weight of BP neural network have been defined. The problem oftrap in the local minimum of the BP neural network has been solved by GA. Theimproved BP model has been applied to calculate and forecast the future series. Incontrast to the normal BP networks, the errors of the calculation are reduced, and thespeed of the convergence is enhanced.2. The stream flow and sediment concentration in a flood process can besimulated by recognizing the pattern of a matrix which contains the information ofthem. Various hydrographs and graphs of sediment concentration in the future havebeen forecasted.3. The improved BP model has been used to calculate and predict of the heightof Tongguan, reservoir sediment scour and deposition, and the distribution ofreservoir sedimentation. The numerical results are consistent with the actualchanging trend. Compared with traditional sediment model, this method need fewerparameters and is faster.4. A random differential equation model for the sediment concentration of thereservoir has been established by using the random differential equation theory. Theresults are reasonable, which indicates that the random differential equation theorycan be used to solve the reservoir sediment problems. The expected values of thesediment concentration calculated by the model are consistent with the actualchanging trend.5. The iteration fuzzy pattern recognition model and GA have been both used inmulti-objective optimal operation of the Sanmenxia reservoir in flood season. Thefast prediction model of the reservoir sediment scour and deposition has been putinto these two models. Perfect solutions are obtained.