引用本文: | 刘剑,李兴源,王成,干华,许立雄.电网分区的主动半监督探测方法[J].电力自动化设备,2018,(6): |
| LIU Jian,LI Xingyuan,WANG Cheng,GAN Hua,XU Lixiong.Power grid partition based on active semi supervised detection method[J].Electric Power Automation Equipment,2018,(6): |
|
摘要: |
分区可以降低电网运行监控和计算分析的复杂度,在实际中得到广泛应用。考虑电力网络演化发展的局域性以及电力系统电压无功的局部性,提出了一种标签节点主动标记的半监督电网分区探测方法:依据电网分区核心节点具有相对较高的连接稠密度且与其他高连接稠密度节点间的距离相对较远的特性,筛选核心节点标记分区标签;依据分区对节点的吸引度,扩散传播分区标签,探测分区组成。为了评价电网分区结果的优劣,借鉴聚类有效性评价的思路,结合电力网络的自身特性,重新定义了silhouettes指标。IEEE 39节点系统算例和某实际电网算例验证了所提方法的有效性和实用性。所提方法可用于电压无功控制、输电断面识别等。 |
关键词: 电力系统 电压影响 电气距离 分区核心 主动学习 标签扩散 聚类分析定 |
DOI:10.16081/j.issn.1006-6047.2018.06.024 |
分类号:TM761 |
基金项目:教育部中央高校基本科研业务费专项资金资助项目(2082604184073) |
|
Power grid partition based on active semi supervised detection method |
LIU Jian1,2, LI Xingyuan1, WANG Cheng3, GAN Hua4, XU Lixiong1
|
1.College of Electrical Engineering and Information Technology, Sichuan University, Chengdu 610065, China;2.Yalong River Hydropower Development Co.,Ltd.,Chengdu 610051, China;3.State Grid Sichuan Electric Power Company, Chengdu 610041, China;4.Chengdu Chengdian Electric Power Engineering Design Co.,Ltd.,Chengdu 610041, China
|
Abstract: |
Network partition can reduce the complexity of operation monitoring, dispatching control and security analysis of power grid and has been widely applied in practice. A semi-supervised detection method with active tag of label nodes for power grid partition is proposed, which considers the localization of power grids' evolution and development and the locality of voltage and reactive power. The core nodes are selected to label the partitions according to their higher connection densities and the relatively larger distances from other nodes with high connection densities, and the labels of partitions are propagated to detect the partitions of the power grid based to the attractiveness of the partitions to the nodes. The silhouettes index is reformulated to evaluate the advantages and disadvantages of power grid partition results based on the idea of clustering validity evaluation and the characteristics of power grid itself. The validity and practicability of the proposed method are verified by the examples of IEEE 39-bus system and a practical power grid. The proposed method is available for voltage and reactive power control, identification of transmission sections, and so on. |
Key words: electric power systems influence of voltage electrical distance center of partition active learning label propagation clustering analysis |