基于电压和功率数据的低压用户相别识别优化建模与应用

Modeling and Application of Phase Identification Optimization for Low-voltage Customer Based on Voltage and Power Data

低压用户相别识别是低压配电网计量拓扑识别的内容之一。首先,描述原问题并对已有研究进行回顾,阐述文中研究特点;按照线路分叉点数量将用户降维到虚拟用户空间;再对虚拟用户空间电压波形采用t分布的随机近邻嵌入(t-SNE)算法,形成进一步降维的特征空间;对于特征空间的电压波形以及聚合功率,提出同时考虑功率平衡和电压时序波形分类的多目标优化模型,使用仅考虑功率平衡的优化问题的解作为初始解。选择2个具有典型特征的城镇居民台区作为试点,发现常见的5类聚类方法均只有1个台区识别成功,且表现差异很大,而所提方法在2个案例中均取得成功,表明其具有一定的稳定性和通用性。

Phase identification for low-voltage customer is one of the contents of metering topology identification for low-voltage distribution networks.First,based on the original problem description and review of existing research,the research characteristics of this paper are explained.The customers are reduced to the virtual customer space according to the number of line bifurcation points.Then,the virtual customer space voltage waveform forms a feature space for further dimension reduction using t-distributed stochastic neighbor embedding(t-SNE)algorithm.For the voltage waveform and aggregated power in the feature space,a multiobjective optimization model considering both power balance and voltage timing waveform classification is proposed,in which the solution of the optimization problem only considering power balance is used as the initial solution.Two urban residential areas with typical characteristics are selected as pilot projects.It is found that only one area is successfully identified by using the common five types of clustering methods and their performance is very different.However,the proposed method identifies both areas,which shows its stability and versatility.