【作者】 李雨；

【导师】 郭生练；

【作者基本信息】 武汉大学 ， 水文学及水资源， 2013， 博士

【摘要】 随着我国各大流域内大型水库群的相继建成,标志着我国水电工作的重点正处于从建设到管理的关键转型期。如何做好已建水库群的管理运行工作,最大程度上发挥水库群的防洪、发电、航运、生态等综合效益,实现水库群的优化调度和经济运行,是当前水库管理面临的首要任务和技术难题,也是目前水文水资源学科研究的热点和重点之一。论文以三峡水库以及清江梯级水库组成的混联水库群为研究背景,开展水库防洪和蓄水优化调度方法方面的研究,主要研究工作及成果如下：(1)回顾了水库调度理论和方法的提出及发展历程,阐述了水库防洪预报调度、优化调度、实时调度的研究进展,全面综述和总结了水库汛末蓄水调度中面临的问题,包括防洪风险问题、泥沙淤积问题以及环境和生态问题。(2)为克服群智能算法中普遍存在的早熟收敛现象,提出了一种新的基于DE算法和PSO算法的并行差分进化粒子群优化算法。前期利用两个分群并行搜索,保证了收敛速度,后期引入收敛因子,判断种群是否陷入局部最优,并利用扩展操作实现种群逃离局部最优区。选用具有不同典型数学特征的10个标准测试函数,进行100次独立仿真实验,验证了算法的优越性。(3)针对三峡和清江梯级水库群防洪补偿联合调度问题,建立了梯级单独和水库群联合防洪优化调度两种数学模型,运用加速收敛的POA算法,引入动态惩罚函数处理约束条件,采用优化、验证、调整三个模块迭代求解。研究结果表明,提出的初始解生成方法,加快了POA算法的收敛速度。选择最不利的1982年为典型年,荆江河段的防洪标准至少可分别提高到118年和136年一遇。(4)针对梯级水库防洪优化调度求解过程中存在的入库洪水过程的随机性,以及优化调度模型求解的复杂性这两方面的问题,提出了基于Copula函数的随机模拟方法,充分考虑了洪水地区组成方式的多样性和随机性,从而能模拟对梯级水库防洪调度最不利的情况；建立了梯级水库防洪优化调度模型,运用基于并行搜索思想的PDP算法,对优化调度模型进行求解。该方法既充分考虑了洪水地区组成的复杂性,又能快速的求解得到最优调洪过程,为梯级水库防洪优化调度研究提供一种新的思路。(5)分析了三峡水库汛末提前蓄水的必要性,在三峡水库现有优化调度和试验性蓄水方案的基础上,研究并拟定了30种汛末提前蓄水方案集,建立了汛末提前蓄水方案优选模型,从防洪风险和综合利用效益两个方面对三峡水库汛末提前蓄水方案进行了优选。建议三峡水库9月1日从145m起蓄,采用分段控制,均匀蓄水的方式,9月30日蓄水至168.0m。(6)应用Copula函数构建了联合分布及条件概率分布,通过随机抽样方法模拟得到三峡水库9月份随机入库流量序列。建立了三峡水库汛末提前蓄水时机优选模型,对9月份提前蓄水方案进行了优选,结果表明：若8月下旬来水为丰水,则9月下旬起蓄,月底均匀蓄水至166m；若来水为平水,则9月中旬起蓄,月底均匀蓄水至166m；若来水为枯水,通过加强实时监测,则可进一步提前至9月上旬起蓄,月底均匀蓄水至166m。更多还原

【Abstract】 As many large reservoirs have been built in China during past thirty years, the main issure of water resources project has begun to transform from construction to management. Therefore, how to strengthen the reservoir group management to maximize the comprehensive benefits of flood control, power generation, nevegation and ecology protection, and to achieve the optimal scheduling and economic operation is the primary task and technical problem for water resources management. In this paper, Three Gorges Reservoir (TGR) and Qingjiang cascade reservoirs were selected as case studies. Optimal scheduel methods of reservoir flood control and refill operation were proposed and discussed. Main research and achievements are summarized as follows:(1) The reservoir operation theories, methods and the development progress were reviewed. Research progress of the reservoir flood control forecast operation, optimal operation and real-time scheduling were presented. In addition, the existing problems durin reservoir refill, including flood control risk, siltation, environmental and ecological problems were summarized(2) In order to overcome the phenomenon of premature convergence which is prevalent in swarm intelligence algorithm, a new algorithm named PDP (Parallel DE and PSO) algorithm was proposed. It is mainly based on the DE and PSO algorithm. The whole population is divided into two parts to keep parallel searching at early period, which can ensure the convergence rate. The convergence factor is introduced at later period to determine whether the populations fall into a local optimum and achieve the populations to escape from the local optimum district by extended operations.10standard test functions with different typical mathematical features were selected and100independent simulation experiments were carried on to verify the superiority of the algorithm.(3) To solve the problem of flood control system of joint operation for the Three Gorges and Qingjiang cascade reservoirs, separate and joint cascade operation flood control models were proposed. The models were solved by a modified progressive optimality algorithm according to the procedure of "optimization-verification-adjustment" and its constraint conditions are treated by dynamic penalty function. The results show that the proposed method of generating the initial solution can accelerate the convergence rate of the POA algorithm greatly. The flood control standard in the Jingjiang River reach can be raised to118-year and136-year design flood respectively based on1982typical flood. (4) There are mainly two difficulties in the process of solving the optimal flood control operation for the cascade reservoirs, which are the uncertainty of reservoir inflow forecasts and the complexity of solution for the optimization model. To solve these problems, a joint distribution function of upper and interval basin floods was established by using copula function and inflow hydrographs of cascade reservoirs were obtained by stochastic simulation method. Furthermore, a flood control optimization model for cascade reservoirs and a new algorithm, named Paralleled DE and PSO (PDP), were proposed. The proposed method not only can consider the variety and randomness of regional flood composition, but also can easier gain the optimum global solution than traditional optimal algorithms, which provides a new approach for flood control operation of cascade reservoirs.(5) The necessity of refill earily of the TGR was analyzed.30different schemes of impounding water in advance were proposed based on scheme of optimal operation and experimental impounding water for the TGR. A procedure coupled a flood control risk module with utilization benefits analysis module is then developed to derive the optimal refill rule. The application results show that the optimal rule is that refill begins on September1with storage level reaching160m on September30linearly.(6) A joint distribution function and conditional probability distribution of this samples using copula was build and inflow series in September were obtained by stochastic simulation method. Furthermore, a refill operation optimization model for the TGR was established to derive the optimal refill scheme. The results show that the optimal refill scheme depends on the reservoir inflow in late August. In the wet year, refill begins on late September with storage level reaching166m on September30linearly; In the normal year, refill begins on middle September with storage level reaching166m on September30linearly; In the dry year, refill begins on early September with storage level reaching166m on September30linearly by strengthening real-time monitoring.更多还原