高量测丢包率下基于共享模糊等价关系的配电网状态感知  

STATE AWARENESS BASED ON SHARED FUZZY EQUIVALENCE RELATIONS FOR DISTRIBUTION SYSTEMS UNDER HIGH PACKET LOSS RATES

作　　者：黄蔓云[1] 马一达 孙国强[1] 卫志农[1] Huang Manyun;Ma Yida;Sun Guoqiang;Wei Zhinong(School of Electrical and Power Engineering,Hohai University,Nanjing 211100,China)

机构地区：[1]河海大学电气与动力工程学院,南京211100

出　　处：《太阳能学报》2024年第3期436-445,共10页Acta Energiae Solaris Sinica

基　　金：国家自然科学基金青年项目(52207090);中央高校基本科研业务费专项资金(B220202003)。

摘　　要：配电网分布式新能源渗透率的不断增加,对实时监测的要求不断提高,大量量测数据在集中传输过程中常常出现延迟、丢包等现象导致配电网状态感知精度降低甚至不可观测。针对上述问题,提出基于共享模糊等价关系的状态感知方法。首先,将采集得到的多时间断面历史量测数据集作为源数据集,实时采集得到的高量测丢包率下的数据作为目标数据集。利用三角形隶属度函数分别将两组数据集写成模糊集的形式。然后采用一种向量间距离度量方法,分别计算得到两组模糊集的模糊等价关系矩阵。在此基础上,采用共享模糊等价关系的聚类方法,将知识从源数据集转移到目标数据集,减小两组数据集特征分布之间的差异。最后,利用深度神经网络对知识迁移后低维特征空间中高量测丢包率对应的数据进行标记,得到配电网实时状态。通过对IEEE标准算例和某实际地市公司配电网算例进行仿真测试,结果表明所提出的基于共享模糊等价关系的配电网状态估计方法能在高量测丢包率下获得准确的配电网实时运行状态,实现不同量测数据缺失程度下的配电网实时监测。The increasing penetration rate of distributed new energy in the distribution network and the increasing demand for real-time monitoring often result in delays,packet loss,and other phenomena in the centralized transmission of a large amount of measurement data,leading to a decrease in the accuracy of distribution network state perception and even unobservability.In response to the above issues,a state awareness method based on shared fuzzy equivalence relationships is proposes.Firstly,the collected historical measurement dataset of multi-time sections is used as the source dataset,and the real-time collected data with high measurement packet loss rate is used as the target dataset.Write two sets of datasets in the form of fuzzy sets using triangular membership functions.Then,a distance measurement method between vectors is used to calculate the fuzzy equivalence relationship matrices of two sets of fuzzy sets.On this basis,a clustering method using shared fuzzy equivalence relationships is adopted to transfer knowledge from the source dataset to the target dataset,reducing the difference in feature distribution between the two sets of datasets.Finally,a deep neural network is used to label the data corresponding to the high measurement packet loss rate in the low dimensional feature space after knowledge transfer,and the real-time status of the distribution network is obtained.Through simulation tests on IEEE standard examples and distribution network examples of a certain actual city company,the results show that the proposed distribution network state estimation method based on shared fuzzy equivalence relationship can obtain accurate real-time operation status of the distribution network under high measurement packet loss rates,and achieve real-time monitoring of the distribution network under different levels of measurement data loss.

关 键 词：新能源 状态估计 丢包 深度学习 共享模糊等价关系 聚类 知识迁移 

分 类 号：V242.3[航空宇航科学与技术—飞行器设计]