基于多智能体遗传算法的配电网故障恢复

发布时间：2018-01-09 04:20

本文关键词：基于多智能体遗传算法的配电网故障恢复　出处：《西南交通大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着配电网的发展,配电网变得越来越复杂,发生故障的概率增大,而配电网故障恢复,能在配网发生故障时,根据配电网的拓扑结构和电气关系及时处理停电故障,恢复非故障停电区域失电负荷的供电,减少停电范围和停电时间等,提高供电质量,减少经济损失。因此,研究快速、准确的配电网故障恢复策略具有重要的意义。本文首先阐述了配电网故障恢复的基本原理、步骤、解决配电网故障恢复这个多目标组合优化问题的多目标进化算法中涉及到的重要定义、一般流程和多智能体遗传算法,为解决配电网故障提供理论依据；然后,针对配电网故障恢复给出了一种快速的拓扑分析和潮流计算方法,运用基于QT的C++实现用GUI (Graphical User Interface)界面显示拓扑和潮流计算结果。结合传统数据库存储和内存存储的优势,解决现有文献不能同时兼顾计算速度和数据维护的方便性的问题。利用SQL关系数据库存储配网的连接关系和节点、支路信息,方便数据维护,提高拓扑重构的灵活性；在拓扑分析前,将数据库中的配网信息加载到内存中形成面向对象的实时数据库,提高计算实时性,再运用广度优先搜索法进行拓扑分析,采用前推回代法进行潮流计算。通过分别对多电源IEEE16节点系统和单电源IEEE33节点系统进行配电网拓扑分析和潮流计算表明：该方法能正确检测配网的连通性、判断是否存在环路或孤岛；具有良好的数据维护性,计算速度快,特别对规模较大配网进行故障分析优势更为显著,为快速恢复故障奠定基础；最后,结合配电网故障恢复问题的特点,给出了一种解决配电网故障恢复的多智能体遗传算法。利用简化的拓扑分析,保证故障支路的编码都为可行解；结合快速非支配排序法和拥挤距离的计算,对多个能量的智能体能量进行排序比较,减少重复比较能量的同时可以很好的权衡多个目标函数；运用计算效率高的擂台赛法则构造得出多目标配电网故障恢复问题的Pareto最优解集。通过采用基于Qt的C++分别对多电源IEEE16节点系统和单电源IEEE33节点系统进行配电网故障恢复仿真,结果表明：该算法有较高的收敛性,并且能够很好的保持种群多样性,避免陷入局部最优解。
[Abstract]:With the development of the distribution network, the distribution network becomes more and more complex, the probability of failure increases, and the distribution network fault recovery, can occur in the distribution network fault. According to the topology structure and electrical relationship of distribution network, power failure can be dealt with in time, power supply of power loss in non-fault blackout area can be restored, power cut range and blackout time can be reduced, and power supply quality can be improved. Therefore, it is of great significance to study the fast and accurate fault recovery strategy of distribution network. In this paper, the basic principle and steps of fault recovery in distribution network are described firstly, and the important definition of multi-objective evolutionary algorithm for solving the multi-objective combinatorial optimization problem of distribution network fault recovery is discussed. General flow and multi-agent genetic algorithm provide theoretical basis for solving distribution network fault. Then, a fast topology analysis and power flow calculation method for distribution network fault recovery is presented. Using QT-based C to implement GUI graphical User Interface. Interface display topology and power flow calculation results. Combined with the advantages of traditional database storage and memory storage. SQL relational database is used to store the connection relation and node of distribution network, branch information, convenient data maintenance. Improve the flexibility of topology reconstruction; Before topology analysis, the distribution network information in the database is loaded into the memory to form an object-oriented real-time database to improve the real-time computing, and then use the breadth-first search method to carry out topology analysis. Through the topology analysis and power flow calculation of multi-power IEEE16 node system and single-source IEEE33 node system, the forward pushback method is used to calculate power flow. This method can correctly detect the connectivity of the distribution network. Determine whether there are loops or isolated islands; It has good data maintainability, fast calculation speed, especially for large-scale distribution network fault analysis advantages are more significant, for the rapid recovery of fault laid the foundation; Finally, according to the characteristics of distribution network fault recovery problem, a multi-agent genetic algorithm is proposed to solve the distribution network fault recovery problem. The simplified topology analysis is used to ensure that the coding of the fault branch is feasible. Combined with the fast undominated sorting method and the calculation of the crowding distance, the multi-energy agent energy can be sorted and compared, which can reduce the repeated comparison energy and balance the multiple objective functions well. The Pareto optimal solution set of the multi-objective distribution network fault recovery problem is obtained by using the high computational efficiency beating race rule. By using QT based on C + +, the optimal solution set of the multi-objective distribution network fault recovery problem is obtained. Simulation of distribution network fault recovery is carried out for multi-power IEEE16 node system and single-power IEEE33 node system. The results show that the algorithm has high convergence and can maintain population diversity and avoid falling into local optimal solution.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM711

【参考文献】