基于SC-DTW距离的低压配电台区拓扑识别方法

Topology Identification Method of Low-voltage Distribution Station Based on SC-DTW Distance

在台区数据采集与存储的过程中,不可避免的会产生数据残缺问题,针对目前多数“变-户”拓扑识别方法使用时间刚性对齐距离来计算电压曲线间相似度,无法应对数据缺失场景下的变-户拓扑识别问题,文章提出一种基于形状上下文动态时间规整(shape context-dynamic time warping,SC-DTW)距离的低压配电台区“变-户”拓扑识别方法,首先,计算电压序列各点处的形状上下文直方图,构建代价矩阵代替原始DTW算法中的距离矩阵;然后搜索累积距离最小的最优对齐路径,得到SC-DTW距离图作为台区各用户电压序列间的相似性度量;最后,进行聚类分析,识别台区“变-户”拓扑关系。所提方法基于时间柔性对齐的距离度量,并引入形状上下文描述子(shape context descriptor,SCD)解决DTW算法存在的病态对齐问题,对数据残缺场景下的变–户拓扑识别问题鲁棒性更高。

In the process of data acquisition and storage in the station,it is inevitable that data will be incomplete.In view of the fact that most of the current"Transformer-User"topology identification methods use time rigid alignment distance to calculate the similarity between voltage curves,which cannot deal with the problem of"Transformer-User"topology identification in data missing scenarios,this paper proposes a"Transformer-User"topology identification method for low-voltage distribution station based on shape context dynamic time warping(shape context-dynamic time warping,SC-DTW)distance.Firstly,the shape context histogram at each point of the voltage sequence is calculated,and the cost matrix is constructed to replace the distance matrix in the original DTW algorithm.The optimal alignment path with the smallest cumulative distance is searched,and the SC-DTW distance map is obtained as the similarity measure between the voltage sequences of each user in the distribution station.Finally,the cluster analysis is carried out to identify the"Transformer-User"topological relationship in the distribution station.This method is based on the distance measurement of time flexible alignment,and introduces the shape context operator(shape context descriptor,SCD)to solve the ill-posed alignment problem of DTW algorithm,which is more robust to the"Transformer-User"topology recognition problem in the case of incomplete data.