【作者】 花蕊；

【导师】 霍利民；

【作者基本信息】 河北农业大学 ， 农业电气化与自动化， 2007， 硕士

【摘要】 网络重构是配电系统运行和控制的重要手段，也是配电自动化系统的重要组成部分。配电网络重构的实现，可以达到降低网损、均衡负荷、提高可靠性等目的。网络重构在理论上是一个复杂的多目标非线性组合优化问题。自上世纪80年代以来，人们对配电网络重构进行了广泛的研究，形成了比较成熟的网络重构的方法和理论。国内外常用的算法，有数学优化理论、最优流模式法(OFP)、支路交换法(BEA)、神经网络算法(ANN)、蚁群最优算法(ACO)、模拟退火算法(SA)、遗传算法(GA)、Tabu算法、专家系统等等。在这些算法中，遗传算法以其全局搜索能力和在其它领域的应用证明了研究遗传算法进行网络重构是有巨大潜力的，这正是本文研究的初衷。根据配电网络的特点，采用变结构耗散网络进行配电网简化分析，并在此基础上实现了网络拓扑分析、网络辐射状和无孤岛判断算法。并确定采用以配电网网损最小作为配电网络重构的目标函数，同时采取事后综合比较开关动作次数的方法确定最优方案，该重构模型灵活实用、易于操作。现有的GA在电力系统中的应用，基本上都是简单地套用基本的GA而忽略了重构问题的特殊性及相应的建模与求解技巧，研究并改进与电力系统相关的遗传编码及遗传操作策略具有重要意义。本文采用了改进的染色体编码策略，即相邻开关在染色体中相邻以及构成同一环路的开关在同一基因块内的编码方法，极大地提高了有效候选解的比例。本文重点阐述了将解决组合优化问题比较有效的单亲遗传算法引用到配电网络重构中。采用了单点移位和令基因块为整个染色体的PGA基因移位操作和修正的PGA基因突变操作。消除了常规遗传算法中对可行性解破坏严重的双亲交叉算子的作用，从而有效的避免了常规算法的缺陷。通过IEEE 16节点配电系统和美国PG&E 69节点配电系统两个标准算例的验证，表明本文提出的方法在配电网络重构的应用是可行的、有效的。最后，在以上理论、算法和算例的研究基础上，以河北省某县城配电网络为原型，采用VC++6.0编程语言，设计并开发了配电网络重构软件。将面向对象技术引入配电网络重构，建立了面向对象的配电网络模型，并对单亲遗传算法进行了有效的封装，使配电网络重构具有灵活性、开放性、易维护性和数据的完整性。该软件作为配电自动化主站端应用系统的一个高级功能模块，现已投入使用。经实践证明：系统人机界面友好，使用简单方便，运行情况良好，运行人员评价较高。更多还原

【Abstract】 This dissertation presents an improved method to study distribution network reconfiguration (DNRC) based on a partheno-genetic algorithm (PGA).Network reconfiguration is not only an important method in circulate and control of distribution system, but also an important part of distribution management system. In theory, network reconfiguration is a complex and many object combination optimization problem. From 1980’ s, people had extensive study on network reconfiguration of distribution system, so many relatively mature methods and theory of network reconfiguration had formed and developed, such as OFP, BEA, ANN, ACO, SA, GA, Tabu and expert system,but all methods have many shortcoming. The ability of global searching and extensive application in other scopes of genetic algorithm had proved that it is attracting goal for us tO study how to apply genetic algorithm to network reconfiguration.On the base of analysis the specialty of distribute system, Structure Variable Dissipated Network (SVDN) is applied as simplified model of distribution network. The DNRC model, in which the objective is to minimize the system power loss, is set up. And the optimal result is gotten by analyzing the system power loss and switches operated times. This model is simple and easy to be implemented.The shortage of classical genetic algorithm in power system was pointed out. The predigestion of networks, the strategy of chromosome encoding and the PGA were proposed. It could be used to reduce the unsolvable code, to enhance the rate of the solvable code and to accelerate the calculating process by the kind of PGA. The paper presents an encoding method of neighboring switches neighboring in the chromosome and switches in the same loop in the same gene. The crossover operation is performed only on the corresponding genes. The mutation and inverse operation is limited within the gene. These strategies greatly reduce the infeasible solutions produced during gene algorithm operations in the application to distribution system reconfiguration. Furthermore, a distribution simplifying method is also proposed to reduce the length of chromosome and enhance the efficiency of the algorithm. The reciprocal value of active power losses is taken to be the fitness function, which will be simple and effective. In particular, the results of two standard test computations show that it could be used to reduce the unsolvable code, to enhance the rate of the solvable code and to accelerate the calculating process by the kind of PGA. At the end of the paper, the solution of true test computations are shown, in which network losing is minimized by distribution reconfiguration. The program is designed by using VC++6.0. The establishment of distribution system model and genetic algorithm is implemented by using Object-Oriented technology (OOT). It makes clear that the proposed algorithm has the effectiveness and practicability更多还原