| 引用本文: | 刘 灏,陈 容,毕天姝,等.基于DBSCAN的配电网同步测量坏数据检测方法[J].电力系统保护与控制,2025,53(17):122-133.[点击复制] |
| LIU Hao,CHEN Rong,BI Tianshu,et al.A DBSCAN-based bad data detection method for distribution network synchronous measurement[J].Power System Protection and Control,2025,53(17):122-133[点击复制] |
|
| 摘要: |
| 配电网环境复杂，配电网同步相量测量装置(distribution network synchronous phasor measurement unit, D-PMU)容易受到干扰而产生坏数据，进一步影响基于测量数据的应用效果。为了提高D-PMU数据质量，提出一种不依赖系统拓扑的基于密度的噪场应用空间聚类(density-based spatial clustering of applications with noise, DBSCAN)的配电网同步测量坏数据检测方法。首先利用基于密度的聚类算法DBSCAN进行异常数据检测。通过轮廓系数和邓恩指数对DBSCAN的聚类结果进行综合评价。利用麻雀搜索算法实现自适应参数调整，解决检测时需要预先处理训练、标记数据的问题。在此基础上，将时间序列聚类的K-Medoids算法和动态时间规整算法相结合，通过衡量不同时间序列之间的相似性，解决了D-PMU 在电气联系较弱时对扰动数据与坏数据的区分问题，增强了数据处理的准确性与噪声环境下的稳健性。仿真和实际数据的测试结果表明，所提方法能有效区分真实扰动数据并准确识别D-PMU坏数据。 |
| 关键词: 坏数据检测 DBSCAN算法 动态时间规整 K-Medoids算法 |
| DOI：10.19783/j.cnki.pspc.241523 |
| 投稿时间：2024-11-14修订日期：2025-03-26 |
| 基金项目:国家电网有限公司科技项目(5100-202199530A- 0-5-ZN，5211DS21N013)“新能源电力系统协同控制保护系统及应用” |
|
| A DBSCAN-based bad data detection method for distribution network synchronous measurement |
| LIU Hao1,CHEN Rong1,BI Tianshu1,ZHAO Dan1,2,ZHANG Yiming1,3 |
| (1. State Key Laboratory of Alternate Electrical Power System with Renewable Energy Sources (North China Electric
Power University), Beijing 102206, China; 2. State Grid Baoding Electric Power Supply Company,
Baoding 071051, China; 3. Huaneng Clean Energy Research Institute, Beijing 102209, China) |
| Abstract: |
| The distribution network environment is complex, and distribution network synchronous phasor measurement units (D-PMU) is susceptible to interference, resulting in bad data that further impacts applications relying on measurement data. In order to improve the data quality of D-PMU, this paper proposes a density-based spatial clustering of applications with noise (DBSCAN) based bad data detection method for distribution network synchronous measurements that does not depend on system topology. First, the DBSCAN is used for anomaly data detection. The clustering results of DBSCAN are comprehensively evaluated using the silhouette coefficient and Dunn index. To address the need for preprocessing training and labeling data during detection, the sparrow search algorithm is used for adaptive parameter adjustment. On this basis, the K-Medoids algorithm for time series clustering is combined with the dynamic time warping algorithm to measure the similarity between different time series, thus solving the difficulty of distinguishing between disturbance data and bad data in D-PMU when the electrical connection is weak. This enhances both the accuracy of data processing and the robustness under noisy environments. Simulation and real data tests show that the proposed method can effectively distinguish real disturbance data and accurately identify bad D-PMU data. |
| Key words: bad data detection DBSCAN algorithm dynamic time warping K-Medoids algorithm |
