| 引用本文: | 陈碧云,吕怡博,梁志坚,等.考虑数据失衡的新型配电网两阶段拓扑辨识[J].电力系统保护与控制,2023,51(21):57-65.[点击复制] |
| CHEN Biyun,LÜ Yibo,LIANG Zhijian,et al.Two-stage topology identification of a new-type distribution network considering data imbalance[J].Power System Protection and Control,2023,51(21):57-65[点击复制] |
|
| 摘要: |
| 在新型电力系统快速发展的背景下，配电网中各类分布式电源、储能、电动汽车、柔性负荷的接入占比不断增加，运行方式日趋复杂多变，其拓扑的精确辨识也更有难度。针对现有配电网量测数据采集周期较长、辨识方法对数据不平衡敏感而导致辨识精度不高的问题，提出了一种两阶段的新型配电网拓扑辨识方法。首先，采用两层堆叠的图卷积网络生成系列标签分类器，再用卷积神经网络提取量测时间序列的特征，并结合多标签分类学习实现第一阶段的初步辨识。其次，对初步辨识获得的初始拓扑中状态为“阴性”(开断)的支路进行全状态空间搜索，并通过潮流匹配模型，筛选出耗散值最小的状态样本，实现“假阴性”二次辨识。最后，在改进的IEEE33节点含新能源配电网络上进行仿真验证。结果表明，所提模型和方法能有效解决数据失衡的问题，并具有更高的辨识精度。 |
| 关键词: 配电网 拓扑辨识 多标签分类 数据失衡 |
| DOI：10.19783/j.cnki.pspc.230503 |
| 投稿时间：2023-05-04修订日期：2023-07-31 |
| 基金项目:国家自然科学基金项目资助(52177085)；广东省基础与应用基础研究基金项目资助(粤桂联合基金-面上项目)(2021A1515410009) |
|
| Two-stage topology identification of a new-type distribution network considering data imbalance |
| CHEN Biyun1,LÜ Yibo1,LIANG Zhijian2,ZHANG Yongjun3,XU Qi1,FU Tianwang1 |
| (1. Guangxi Key Laboratory of Power System Optimization and Energy Saving Technology (Guangxi University),
Nanning 530004, China; 2. School of Electrical Engineering, Guangxi University, Nanning 530004, China;
3. School of Electric Power, South China University of Technology, Guangzhou 510640, China) |
| Abstract: |
| Along with the rapid development of new-type power systems, the proportion of access to various types of distributed power sources, energy storage, electric vehicles, and flexible loads in the distribution network is increasing. The operation mode is becoming more complex and changeable, and the accurate identification of their topology is more difficult. There is a problem in that the data collection cycle of distribution network measurement is long and the identification method is sensitive to data imbalance, resulting in low identification accuracy. Thus this paper proposes a two-stage topology identification method of a new-type distribution network. First, a two-layer stacked graph convolutional network is used to generate a series of label classifiers, and then the convolutional neural network is used to extract the features of the measured time series, and the preliminary identification of the first stage is realized by combining multi-label classification learning. Second, the branches with a state of "negative" (breaking) in the initial topology obtained by preliminary identification are searched for the whole state space. The state samples with the smallest dissipation value are screened out through a power flow matching model to achieve a secondary identification of "false negative". Finally, simulation verification is carried out on the modified IEEE33 node power distribution network with new energy. The results show that the proposed model and method can effectively solve the data imbalance problem and have higher identification accuracy. |
| Key words: distribution grid topology identification multi-label classification data imbalance |
