【摘要】 智能电表的广泛普及和高级测量体系（advancedmeteringinfrastructure,AMI）的建立为分析配电网运行情况提供了大量监测信息与测量数据，而台区用户的相位信息变动又为准确掌握台区运行情况带来难题。针对台区用户的相位识别问题，提出了一种基于用户电压数据的t分布随机邻接嵌入（t-distributed stochastic neighbor embedding,t-SNE）特征提取及放射传播（affinity propagation,AP）聚类算法的相位识别方法。先对提取出的用户电压数据进行Z-score数据标准化处理，由t-SNE降维提取出数据特征，再采用放射传播聚类算法对用户进行相位识别。选取某市2个小区进行算例分析，采用评价指标比较了不同识别方法的识别效果，并分析了数据采集频率和计量误差对识别效果的影响。实际台区算例分析验证了所提方法的准确性，说明所提方法能够有效解决台区用户相位识别问题。更多还原

【Abstract】 The widespread popularity of smart meters and the establishment of advanced measurement infrastructure(AMI) provide a large amount of monitoring information and measurement data for the analysis of the operation of distribution networks, while the change of phase information of users in the station area brings difficulties to the accurate understand of the operation of the station area. Aiming at the problem of phase recognition of users in the station area, a phase recognition method is proposed based on tdistributed stochastic neighbor embedding(t-SNE) feature extraction and affinity propagation(AP) clustering algorithm of user voltage data. Firstly, the extracted user’s voltage data is processed by Z-score data standardization, and the data features are extracted by t-SNE dimensionality reduction. And then phase identification for the user is made with radial propagation clustering algorithm.Two districts in a city are selected for case study. The recognition effects of different recognition methods are compared using evaluation indicators, and the effects of different acquisition frequencies and different measurement errors on the recognition effects are analyzed. The accuracy of the proposed method is verified by the actual cases, which shows that the proposed method can effectively solve the problem of user phase identification in the station area.更多还原