Inflow Analysis And Optimal Dispatching Of Jinpen Reservoir In Shaanxi Province

The Heihe diversion project constructed for solving the problem of water supply in Xian City is the basic guarantee to sustainable and high-speed development of Xian's economy. And the operation state of the Jinpen reservoir, which is the main water source of the diversion project, will make direct influence on the water supply of the city.In this paper, combined with the practical situation of water supply of Xian City and the Jinpen reservoir, the inflow characteristics and water requirement state of every water supply object are analyzed, the optimal operation model of this complex utilization reservoir is developed with consideration of its multi-objective and nonlinear features and several models are used to make the optimal dispatching of the reservoir, then the reservoir optimal operate function is established to direct the reservoir operation. The main research content and results are as follows:(1) Based on detailed description of current research trend and development in this field and analysis of the basic data of the reservoir and the watershed, the aims, contents and methods of this study are put forward.(2) The method of time series analysis and the stochastic models are introduced, and the yearly and monthly runoff series stochastic models at the Heiyukou gauging station in Heihe River are established and checked.(3) Taking into account the fuzzy characteristics existed in selecting of the representative year, the fuzzy clustering method is adopted to the cluster analysis of year runoff data, and the cluster center vectors standing for the type of year, that is, abundant, moderate and deficient, are obtained. Then for the new given data, the fuzzy synthetic evaluation method can be used to determine its type according to the derived cluster center vectors.(4) Based on the analysis of the inflow characteristics and water requirement state of every water supply object of the reservoir, the optimal operation model of this complex utilization reservoir is developed with consideration of its multi-objective and nonlinear features. The model is solved by the three methods of dynamic programming, the progressive optimization algorithm and the genetic algorithm, and the calculation results obtained from different methods are compared.(5) The basic theory and realizing techniques of genetic algorithms are systematically introduced, and the shortcomings of GA are analyzed, then some improving methods are presented. Premature convergence is still the prominent problem in genetic algorithms. Some excellent individuals or genes are lost due to premature convergence, which causes local optimum. A parameter adaptive genetic algorithm based on the entropy estimating is advanced in this paper. The parameter can be adjusted adaptively by incorporating population entropy into the algorithm to provide a quantitative measure of population diversity of individuals in the population and new individuals can be introduced into the population by a big-mutation operation when the individuals in the population tend to be consistent. The case study shows that the algorithm can alleviate the problem of premature convergence as well as improve the efficiency and scope of searching.(6) The reservoir operation function is established utilizing the optimal dispatching results. The state variable and decision variable of the operation function is discussed, and the stepwise regression method is used to derive the operation function. Because of the nonlinear features of the function haven't been reflected in traditional regression methods, the back-propagation neural network model is introduced to establish the operation function. And the operation functions established by different method are checked using the unused data.