【作者】 杨丽君；

【导师】 卢志刚；

【作者基本信息】 燕山大学 ， 电力电子与电力传动， 2010， 博士

【摘要】 电网互联规模的扩大,满足了现代社会对电力供应日益增长的需求,同时网络结构的日趋复杂也给电网运行带来很多潜在的威胁,一旦发生大面积停电事故,所造成的社会影响和经济损失将极其严重。作为电力系统安全防御的重要措施之一,研究电力系统大停电后安全、快速、有序地恢复系统供电,是减少大停电损失、加快系统恢复的重要手段,也是构建智能电网的重要内容,具有重要的理论意义和社会价值。大停电后的电力系统恢复是一个复杂的控制和决策过程,具有多目标、多阶段、多方参与的属性,通常划分为三个阶段:黑启动阶段、网架重构阶段和负荷恢复阶段。利用各种优化方法实现大停电后系统恢复的决策分析是必要的手段。根据大停电后系统恢复进程的三个阶段,本文进行了以下研究工作:(1)提出一种基于支持向量机理论的过电压预测方法,将其过电压的预测结果与EMTP的计算结果、及神经网络的预测结果进行比较。结果表明采用支持向量机理论能够实现过电压快速、准确预测,能够满足黑启动初期空充长输电线路过程中对可能出现的过电压的校验需要。(2)提出一种利用复杂网络理论评估电网节点重要性的方法。首先将电网抽象成一种赋权值的复杂网络,采用复杂加权网络的节点凝聚方法计算电网中各节点的重要度指标,客观上评价出较为重要的网络重构节点。根据电源和负荷的实际重要性进行适当调整,从而确定进行网络重构的目标节点。在此基础上,利用Kruskal最小生成树算法,计算出覆盖网络所有节点的最小生成树拓扑。通过适当地删减树枝,构造一个覆盖所有重要节点的局部最小生成树,可将此局部最小生成树拓扑作为重构主干网架的一个目标网络。(3)针对网架重构阶段的目标,提出一种基于蚁群优化算法结合最短路径算法的主干网架重构方法,优化主干网架同时,亦优化主干网架的各节点路径投入顺序。该算法过程分为两个层次,外层计算利用蚁群算法实现节点投入顺序优化。在各重要节点之间建立广义路径,动态调整已供电区域路径的线路权值。内层利用最短路径算法计算每一待投入节点与已供电区域间的最短路径和距离,并为蚁群算法信息素更新提供依据。算例表明该算法具有较高的计算效率,收敛速度快。(4)针对大规模负荷恢复阶段的任务,提出一种改进蚁群算法实现负荷恢复阶段网络优化的方法。保留适当系统备用容量的前提下,通过扩展潮流计算方法,计算当前时段系统可恢复的最大负荷量,以此负荷恢复量最大为目标函数,利用改进蚁群算法实现网络优化。蚂蚁选择路径的过程采用随机拓扑的搜索策略,能扩大蚂蚁搜索范围,确保有更多机会搜索到全局最优解,且保证所有恢复路径的连通性。为提高方案可行性验证速度,引入蚁群搜索的模式记忆功能,记录已搜索验证过的部分较好恢复方案,对目标值较小的方案不再进行潮流计算验证,避免大量反复验证恢复方案可行性的问题,提高了整个算法的计算效率。算例表明利用改进蚁群算法进行负荷恢复的优化过程快速、有效,适合于负荷恢复的各个阶段。更多还原

【Abstract】 The interconnection of power grids meets the growing demand for electricity in modern society, also leads to more complex network structure which brings lots of potential threat to the power system operation. The effects of wide area outages in such huge power system may be quite more severe too. Therefore, as an important defense measure, the research on the restoration of the power supply safely, promptly and orderly after wide area outage which caused by faults is of great theoretical and practical significance. It will help to speed up the system restoration and reduce the loss. It’s also an important part in the construction of the smart power network.The problem of restoration after wide area service interruption is a complex decision-making and control problem. It can be described as a multi-objective, multiple-participant, multi-stage, combinatorial, nonlinear optimization problem with constraints. Using appropriate optimization algorithms to make decision in the restoration process after wide area outage is quite essential. According to the three stages of a general restoration process, this thesis focuses on the following issues:1. A method to predict the over-voltage that may appear during the process of energizing the no-load long transmission lines by using the Support Vector Machine theory is presented. The comparison with the Neural Network’s prediction results and EMTP calculation results shows that the method to predict over-voltage with the Support Vector Machine theory works well, and can meet the prediction requirement for speed and accuracy.2. A method to evaluate the node importance in power system grid is proposed. Describe the power grid in a weighted complex first, and then an improved node contraction and node importance evaluation method in weighted complex network is adopted to evaluate the node importance in weighted power system network. Determine the important nodes to reconstruct in current stage after appropriate adjustments based on the nodes actual importance and the importance assessment results above. Then by using the Kruskal algorithm to construct the minimum spanning tree of the power network, through the appropriate branches cut , a local minimum spanning tree that covering all the selected important nodes can be constructed as the skeleton- network.3. An Ant Colony algorithm combining with the Shortest Path algorithm is adopted to optimize the skeleton- network structure, give the path and node restoration sequence too. The combined algorithm consists of two layers. The outer calculation solves the nodes sequence by Ant Colony algorithm. The shortest path algorithm in the inner calculation provides the distance and path between each important node and the equivalent power supply area. The distance is also the basis for updating Pheromone in Ant Colony algorithm. Examples show the effectiveness of the method.4. An improved Ant Colony algorithm is used for the network optimization in the load restoration phase. The maximum amount of recovery load in current step can be calculated by an extended power flow method on the basis of system reserve capacity. Then an improved Ant Colony algorithm which combined with a random topology search strategy is used to optimize the network for load recovery. Improved pheromone update strategy expand the searching scope of the proposed algorithm, ensure a greater opportunity for the global optimal solution. By memorize some better recovery modes searched, the time used to verify the recovery schema by time-consuming power flow calculation can be greatly reduced. Several examples show the effectiveness of the proposed algorithm for load recovery optimization. The proposed algorithm is suitable for all stages in load recovery optimization.更多还原