【作者】 王赢；

【导师】 周建中；

【作者基本信息】 华中科技大学 ， 系统分析与集成， 2012， 博士

【摘要】 流域梯级水库群优化调度一直是学术和工程界研究的前沿和热点问题。受流域气象、水文过程、发电控制、电网运行以及生活生态用水的影响,呈现出高维、非凸、非线性、耦合性强的特性。随着我国水电能源的大规模迅速发展,一批在建水库相继投产运行,梯级水库群规模日益庞大,结构日趋复杂,传统仅考虑孤立水库的单目标调度方式已经无法满足现阶段流域梯级大型水利枢纽综合优化调度运行的需要。为此,亟需综合考虑流域梯级上下游电站之间的复杂水力、电力耦合关系,从流域大规模水电能源系统整体优化的角度出发,针对梯级水电站群联合优化调度的工程需求提出先进的决策理论与方法,统筹协调流域梯级上下游电站群各单位各部门的综合利益和冲突,从而提高梯级流域梯级水电系统的运行管理水平,实现流域水能资源的充分利用。因此,本文以梯级水库群联合优化调度为研究背景,结合系统工程理论和现代智能优化方法针对梯级水库群运行中防洪、发电的主要问题进行研究,并将成果应用于三峡梯级水库优化调度的问题中。本文研究工作及研究方法包括：(1)针对传统算法求解水库防洪调度问题存在维数灾、计算效率低的缺陷,提出了一种混沌文化粒子群算法求解水库防洪调度优化问题。该算法以粒子群算法为主要的进化搜索方法,结合文化算法的计算框架对种群的进化过程进行指导,实现了文化算法通信协议的操作方式。此外,为增强算法的收敛效果,在信念空间中设计了混沌局部搜索策略。结合种群和信念空间的双重进化,一方面加快算法的收敛速度,另一方面有效防止算法早熟收敛。将混沌文化粒子群算法应用于测试函数以及三峡水库防洪优化调度的结果表明,该方法相比传统的防洪调度方法拥有更好的全局搜索能力,为求解梯级水库群防洪优化调度问题提供了一种有效方法。(2)由于梯级水库群优化发电调度问题需要考虑更多复杂的等值与不等值约束,而传统的进化算法缺少对这类复杂约束进行处理的能力。针对这一问题,建立了-种改进约束处理的自适应混沌粒子群算法,该方法将发电调度的初末水位约束转换为水量约束,并围绕水量约束的调整过程提出了两种处理水库优化调度问题中的等值约束处理方案,一是将水量约束分配至随机时段,直到水量约束满足,二是将水量约束中超出或者不足的部分平分成指定份数,然后每一份水量按照最优目标值原则分配。基于上述两种方案的混合约束处理策略设计了算法求解过程中调度方案的约束处理方法。此外,引入一种混沌搜索策略,以改进算法的搜索效率。将改进约束处理的自适应粒子群算法应用于三峡梯级枢纽中长期发电调度模型以及扩展至水火电联合优化仿真系统的求解,结果证明了该算法求解梯级水库群联合优化发电调度问题的有效性,且拥有较高的求解精度和效率,为流域梯级水库群的发电优化调度提供了有力的工具。(3)针对梯级水库群综合开发利用的多目标工程需求,提出了多目标改进约束处理的自适应混沌粒子群算法对该类问题进行求解。该方法结合多目标算法的相关理论,设计了一种多目标问题的种群规划方法。将算法的种群设计为包含局部种群、局部种群历史最优解集、优秀个体集三个部分,减少了计算过程中重复的合并种群的操作,提高算法效率。算法引入快速非支配排序的方法对算法种群构建非支配集,提出了基于欧氏距离判断的个体筛选策略对进化过程中的调度方案集进行优选。建立了一种多目标粒子群算法中个体历史最优解和种群全局最优解的选取策略,将等值约束处理方法扩展至适用于多目标问题的求解。测试函数以及三峡梯级多目标防洪、发电调度的应用结果表明该方法收敛精度高,分布性好,能够为三峡水库优化调度问题的求解提供数据支持。(4)将流域梯级水库群优化调度相关理论与方法结合工程实践,与现代计算机技术结合,构建梯级优化调度系统以指导水库群调度工作的实际工程运行。结合B/S系统架构的优点,设计了SS框架,并应用于三峡梯级水库群优化调度系统的开发。系统结合Struts2框架建立了系统的表示层,引入DWR技术优化JS对后台的调用。基于Web服务,开发了分布式的业务应用层,实现了跨平台、跨语言的模型算法调用,提出了简化数据层的框架,避免了传统数据层框架的复杂配置,优化了数据的访问。此外,通过优化三峡梯级优化调度系统的类设计方式,结构化系统中各种类和对象的设计,提出了一种考虑非物理意义对应关系的数据库设计模式,提高数据的查询效率。为提高系统的开发效率,设计了数据层代码生成方法以优化数据层开发过程。针对面向对象环境下的系统开发,提出了两种通用数据折半查找方案,并分析了两种方案的使用范围。系统的设计简化了开发过程,增加了系统可维护性,提高了调度工作的整体应用效果,为流域梯级优化调度提供了一个快速高效的工作平台。更多还原

【Abstract】 The scheduling of cascade hydroelectric stations is a famous problem that has drawn a wide attention by many scholars. By the effective of meteorological, hydrology, power generation, the work of electric network and the need of water for live, it has proved to be a high-dimension, non-convex, nonlinear and strong-coupling problem. As the great development of water resources in our nation, more and more reservoirs have been or are in contribution. The hydro systems have gone more and more complex. The tradition methods only take single reservoir or single object in consideration, however, nowadays, to raise the synthetically benefit of cascade hydroelectric stations, the relation of water resource and power between up and down reservoir should also be paid attention to. What’s more, the optimal thesis and method is also necessary. Only the profit and conflict of different department are coordinated that the management of systems can be raised, even the fully utilize of the water resources. In this case, my work focuses on the optimal operation of cascade hydroelectric stations. To study the problem of flood control and power generation in hydro scheduling, my work imported the thesis of system-engineering and the modern optimization algorithm. Finally, the result of this work is applied to the operation of Sanxia project in our nation. The main work is described as:(1) For the reason that the tradition method exhibit the drawback of "diverse disaster" and high computation while applied to reservoir scheduling problem. In my work, a chaotic cultural particle swarm optimization is proposed to solve this problem. This method takes the particle swarm optimization (PSO) as the main searching method. The framework of cultural algorithm (CA) is applied to direct the evolution process of population. Also a communication protocol of CA, which consists of accept and effect operation, is also accomplished. To strengthen the effective of proposed algorithm, the chaotic local search is imported in the believe space. With the evolution of two spaces, not only the convergence speed of algorithm is accelerated, but also the premature can be prevented. The result of applying proposed method to the test function and the flood control of Sanxia project proved that, while compared with tradition method, the proposed algorithm has greater searching ability, which is an effective method for the flood control problem of cascade hydroelectric stations. (2) The scheduling problem of cascade hydroelectric stations consist more complex equal and unequal constraints. However, the tradition methods lack of strategy to deal with these constraints. So an improved-constraint self-adaptive chaotic particle swarm optimization is proposed to overcome this drawback. In this method, the initial and final water level constraints are converted to another equality constraint, the quantity of water constraint. Then two strategies are proposed to handling this constraint. One is that by allocating the difference water of equality constraint to different intervals randomly, another is that dividing the difference water of equality constraint into small parts, then allocates the part water by the rule of better object functions until the equality constraint is meet. What’s more, a new chaotic local search strategy is applied to improve the searching efficiency of proposed algorithm. While applied the method to solve the problem of optimal generation scheduling in cascaded hydroelectric stations and the test system of hydro-thermal scheduling, the proposed algorithm shows high efficiency and convergence. Obviously that proved the method is feasible for generation scheduling in cascaded hydroelectric stations.(3) To fully utilize the various functions of cascaded hydroelectric stations, my work has proposed a multi-object improved-constraint self-adaptive chaotic particle swarm optimization to solve the multi-object problem on the operation of cascaded hydroelectric stations. This method is based on the thesis of multi-object algorithms. To reduce the procedure of merging population in evolution, a combination population is devised, which includes three parts:the local population, the best solutions for local population and the better solutions. The fast dominated sort method of NSGA-II is imported to build the non-dominated set. Then the strategy of selecting individuals which proposed in SPEA2 is applied to select the excellent individuals. A way of selecting the history and overall best solution for multi-object problem is also defined. What’s more, the constraint handling method proposed in previous algorithm is improved to adapt to multi-object problem. The result of applied proposed algorithm to solve multi-object test functions, flood control and power generation problem in Sanxia project shows that the proposed algorithm has the better solution in both convergence and diversity, which provide a better data support for the problem of operation in Sanxia project. (4) By integrating the optimal thesis and method in an operation system that these achievements can direct the operation procedure of cascaded hydroelectric stations directly. In that case, a B/S based structure named SS framework to design the system of operation. The Struts2 framework is applied to the display layer of system. To optimal the procedure that JS operates the method in server, the DWR technique is adopted, which makes the configuration of system much easier. The Web Service technique is applied to the design procedure of business layer that the feature in different models and algorithms, cross platform and language, is achieved. The simple data layer is proposed to avoid the complex configure of tradition data layer framework. It’s more important that the reference procedure of data is optimized. A way of designing class for operation system is defined. The design pattern of considering the relation that exists in unrelated object is applied to adequate the efficiency of visiting data. What’s more, a method of generating code in simple data layer is proposed, which make the development process much faster. Two object oriented based half search method is also raised. And the range of application of two methods is also shown. The application of operation system has achieved good results, that a fast and convenient work platform is provided.更多还原