Research On Uncertainty Analysis And Multi-attribute Decision Models Of Cascade Reservoirs' Operation

In order to address the shortage of fossil energy,air pollution,greenhouse effect and other problems,China attaches great importance to and actively promotes the development of clean,renewable,abundant and widely-distributed green energy such as hydropower.Reservoirs play an important role in flood prevention,disaster alleviation,hydropower generation and water supply.In recent years,with the gradual completion of cascade reservoirs in China's major basins and under the impact of global extreme climate change,the relationship among flood control,power generation,water supply and other sectors is getting more and more complicated.Therefore,carrying out the optimal reservoir operation and management under uncertain conditions and seeking more practical optimal operation schemes have become an important issue to be solved by water conservancy and power sectors.In this thesis,following the tack of "uncertainty reduction—uncertainty quantification-operation risk estimation considering uncertainty-uncertain multi-attribute decision-making" and using the theories and methods in mathematical statistics,risk analysis,operational research and Copula functions,an in-depth study is conducted on flood inconsistency analysis,reservoir inflow forecast error simulation,multi-attribute decision-making and other realms.The main achievements are as follows.(1)Joint analysis of multiple variables in inconsistent floods based on Copula functions.To address the problem that the traditional flood frequency analysis does not consider flood inconsistency,the P-? mixed distribution of flood volume variables and the Von Mises distribution of flood occurrence time variables are established.On this basis,the joint distribution of flood volume variables and flood occurrence time variables is established by using Copula functions.Taking the Jinping I reservoir's inflow flood as an example,the feasibility and effectiveness of this method is proved by the calculation of joint exceeding probability distribution,conditional exceeding probability density,etc.(2)Stochastic model of reservoir inflow forecast errors and its application.In order to effectively improve the accuracy of formulating reservoir operation schemes under the condition of quantifying inflow forecast errors,a stochastic model of reservoir inflow forecast errors at multiple prediction moments is developed,based on the Gaussian Mixture Model with good adaptability which can more accurately describe the inflow forecast error at a single prediction moment and the high-dimensional meta-student t Copula function which can integrate various types of edge distribution.Taking the Jinping I reservoir's daily inflow as an example,the forecast errors at multiple prediction moments are randomly simulated to verify the feasibility and validity of the model.(3)Risk estimation for the short-term power generation operation of cascade reservoirs considering multi-dimensional inflow forecast errors.Taking the cascade system with two reservoirs as an example,the historical inflow forecast errors are classified and the uncertainty probability is defined as one of the benefit indicators of operation decision-making.The short-term optimal power generation operation model of cascade reservoirs considering multi-dimensional inflow forecast errors is established and is solved by optimization algorithms to obtain the optimal operation process.Based on the idea of stochastic simulation of inflow forecast errors,the possible future inflow is obtained,and then simulation scheduling is carried out to obtain the estimated values of risk indexes.Compared with the operation scheme made according to the forecast inflow,the model takes the prediction error into account and thus makes more practical schemes.(4)Multi-dimensional interval number decision model based on Mahalanobis-Taguchi System with grey entropy method and its application.In view of the problems in interval number decision-making(e.g.,how to reduce information loss to enhance decision accuracy and how to rank interval numbers),the grey entropy method is improved by the Mahalanobis-Taguchi System which has few orthogonal tests,obtains abundant information and uses the Mahalanobis distance to better reflect the correlation between indexes.Then the improved grey entropy method is integrated with the Mahalanobis-Taguchi System to develop a multi-dimensional interval number decision model.The model is applied to the selection of the Pankou reservoir's multi-objective optimal operation schemes and the Three Gorges cascade reservoirs' optimal flood control operation schemes.Via analysis of the decision results compared with other methods,the merits of the model are verified.