| 引用本文: | 沈 赋,张 微,徐潇源,等.基于随机森林和最大互信息系数关键特征选择的配电网拓扑辨识研究[J].电力系统保护与控制,2024,52(17):1-11.[点击复制] |
| SHEN Fu,ZHANG Wei,XU Xiaoyuan,et al.Topological identification of distribution networks based on key feature selection using RF and MIC[J].Power System Protection and Control,2024,52(17):1-11[点击复制] |
|
| 摘要: |
| 随着高比例分布式能源(distributed generation, DG)的接入，配电网的拓扑变化更加频繁。针对含DG的配电网拓扑辨识所需量测特征多、辨识准确率低的问题，提出基于随机森林(random forest, RF)算法和最大互信息系数(maximal information coefficient, MIC)关键特征选择的配电网拓扑辨识方法。首先，考虑风光出力的不确定性和相关性，基于Frank-Copula函数得到典型风光出力场景，与配电网不同拓扑相结合构建数据集。然后，根据RF和MIC进行特征选择，筛选出对拓扑辨识最重要且不含冗余信息的关键特征。最后，利用蝙蝠算法(bat algorithm, BA)优化BP(back propagation, BP)神经网络模型对配电网拓扑模型进行辨识。通过IEEE 33节点配电网和PG&E 69节点配电网进行仿真分析，验证所提模型的可行性。 |
| 关键词: 配电网 拓扑辨识 不确定性 相关性 特征选择 |
| DOI：10.19783/j.cnki.pspc.240088 |
| 投稿时间：2024-01-19修订日期：2024-05-21 |
| 基金项目:国家自然科学基金项目资助(52107097)；云南省应用基础研究计划项目资助(202101BE070001-061， 202201AU070111)；昆明理工大学高层次人才平台建设项目资助(KKZ7202004004) |
|
| Topological identification of distribution networks based on key feature selection using RF and MIC |
| SHEN Fu1,ZHANG Wei1,XU Xiaoyuan2,WANG Jian1,FU Yu1,YANG Guangbing1,ZHAI Suwei3 |
| (1. Faculty of Electric Power Engineering, Kunming University of Science and Technology, Kunming 650500, China;
2. Department of Information and Electrical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China;
3. Electric Power Research Institute, Yunnan Power Grid Co., Ltd., Kunming 650217, China) |
| Abstract: |
| As the integration of a high proportion distributed generation (DG) into the power grid increases, topological changes in the distribution network become more frequent. A method for the topological identification in the distribution network based on the random forest (RF) and the maximal information coefficient (MIC) for key feature selection is proposed to address the issues of high measurement feature requirements and low identification accuracy in that identification for such networks containing DG. First, considering the uncertainty and correlation of wind and solar power output, typical scenarios of wind and solar power output are obtained based on the Frank-Copula function, and combined with different distribution network topologies to construct a dataset. Then, feature selection is performed using RF and MIC to identify the most important and key non-redundant features for topological identification. Finally, the bat algorithm (BA) is employed to optimize a back propagation (BP) neural network model for identification. Simulation analyses are conducted on the IEEE33 and the PG&E69-bus distribution networks to validate the feasibility of the proposed model. |
| Key words: distribution network topology identification uncertainty correlation feature selection |
