| 引用本文: | 郭上华,王钢.基于多粒度聚类和多元特征统计的低压配电网拓扑识别与监测[J].电力自动化设备,2023,43(6): |
| GUO Shanghua,WANG Gang.Topology identification and monitoring of low-voltage distribution network based on multi-granularity clustering and multivariate characteristic statistics[J].Electric Power Automation Equipment,2023,43(6): |
|
| 摘要: |
| 针对低压配电网中用户的户变关系及相位信息常存在错误且变动较为频繁的现象，提出了一种基于多粒度聚类和多元特征统计的低压配电网拓扑识别与监测方法，包含拓扑聚类识别和拓扑统计监测2个阶段。在拓扑聚类识别阶段，基于用户电压波动曲线的α-峭度和α-偏度提取数据粗粒度空间特征，采用密度峰值聚类算法识别所有用户的户变关系；在数据细粒度特征空间，通过考虑延迟效应的动态时间弯曲距离算法优化密度计算过程，实现相位关系的精确聚类。在拓扑统计监测阶段，基于多粒度聚类结果，采用邻域保持嵌入算法建立多元统计监测模型，实现新增用户或拓扑有变化的个别用户的拓扑快速识别。实际算例的分析结果验证了所提方法的有效性。 |
| 关键词: 配电网 拓扑结构 户变关系 相位识别 密度峰值聚类 多元特征统计 粒计算 |
| DOI：10.16081/j.epae.202211008 |
| 分类号:TM727 |
| 基金项目:国家自然科学基金资助项目(51477057) |
|
| Topology identification and monitoring of low-voltage distribution network based on multi-granularity clustering and multivariate characteristic statistics |
|
GUO Shanghua1,2, WANG Gang1
|
|
1.School of Electric Power Engineering, South China University of Technology, Guangzhou 510640, China;2.Henan Xuji Instrument Co.,Ltd.,Xuchang 461000, China
|
| Abstract: |
| In view of the errors and frequent changes in the consumer-transformer relationship and phase information of users in low-voltage distribution network, a topology identification and monitoring method of low-voltage distribution network based on multi-granularity clustering and multivariate characteristic statistics is proposed, which includes topology cluster identification stage and topology statistical monitoring stage. In the topology cluster identification stage, the coarse-grained spatial features of data are extracted based on the α-kurtosis and α-skewness of the user voltage fluctuation curves, and the consumer-transformer relationship of all users are identified by using the density peak clustering algorithm. In the fine-grained feature space of data, the dynamic time warping distance algorithm considering the delay effect is used to optimize the density calculation process and achieve the accurate clustering of phase relationship. In the topology statistical monitoring stage, based on the multi-granularity clustering results, the multivariate statistical monitoring model is established by using the neighborhood-preserving embedding algorithm to realize rapid topology identification of new users or individual users with topology changes. The effectiveness of the proposed method is verified by the analysis results of a practical example. |
| Key words: distribution network topology structure consumer-transformer relationship phase identification density peak clustering multivariate characteristic statistics granularity calculation |
