Computational Statistics Methods And Their Applications To Risk Management In Reservoir Flood Utilization

A framework of risk management in reservoir flood utilization was introduced after the basic concept, theory and application of flood utilization, risk management and computational statistics were reviewed. Risk identification, estimation and standard of the flood risk which may occur in flood utilization by adjusting limiting level during flood season for reservoir in planning and design phase were studied intensively in this dissertation.First, Bootstrap was applied to estimate statistic, including mean, coefficient of deviation, coefficient of skew and quantile, standard deviation and confidence interval, while population sampled from Pearson III distribution. The contrast demonstrated Bootstrap was more effective than formula calculation. Then Bootstrap, Copula function and Monte Carlo were employed to estimate uncertainty in water-level capacity curve, water-level area curve, discharge, flood frequency and flood design and stochastic simulation model parameter and structure. Thus the key uncertainty factor can be identified by estimating the degree these uncertainty influence the water level or Risk Rate. The estimation of flood peak and volume during flood design and stochastic simulation model parameter were the key uncertainty and computational statistics such as Bootstrap, Copula function and Monte Carlo were effective in risk identification.Four stochastic flood process simulation models were employed in Risk Rate estimation. The results showed that every model hold dependence structure between peak and volume and the Risk Rate is reasonable. These illustrated that dependence structure between peak and volume can be used to simulate flood process. Rank correlation can be adopted to simulation dependence structure. Moreover Rank correlation hold many good character such as distribution free, "scale-invariant" (remain unchanged under strictly increasing transformations of the random variables) and can descript nonlinear dependence structure. So it can be employed to descript the population is not normal distribution and dependence structure is nonlinear.A complex random event was constructed at first that if the water level exceed the flood standard in either flood season, the flood standard was exceed. Then recurrence interval formula was derived in order to define flood design standard of each flood season. When flood design was used to descript flood standard the relation among flood standard in every flood season and flood standard before flood division can be illustrated by the joint distribution of flood peak or flood volume, determined by the domination of flood peak or flood volume reservoir, in all flood season. Under the assumption that flood peak or flood volume between flood seasons is independent, Combination frequency method can be used to descript the relation; while they are associated, Copula function was applied to construction of the joint distribution of flood peak or flood volume. In application part Combination frequency method and different Copula function were both adopted to construction flood peak and flood volume flood joint distribution at Panjiakou reservoir whose flood season were divided into two period. Through the comparison among the results of above method, Flood design standard under the independent assumption is lower than the results under the dependent assumption when the two flood standards are the same, but results among different Copula function were slightly different. The combination of flood standards between the two assumptions was totally different, while among different copula function were slightly different. So particular emphasis should be placed on the dependence structure between flood peak or flood volume among two flood seasons.