基于多特征符号聚合近似和层次聚类的户变关系识别方法

Identification Method for User-transformer Relationship Based on Multi-feature Symbolic Aggregate Approximation and Hierarchical Clustering

针对低压配电台区拓扑档案中可能存在的户变关系异常问题,文中提出了一种基于多特征符号聚合近似(MF-SAX)和层次聚类的户变关系识别方法。首先,运用符号聚合近似表达方法将用户电压时间序列转化为字符串序列,并引入电压波动系数和电压变化趋势两个附加参数对其特征表达进行强化。然后,基于编辑距离生成用户电压曲线相似性矩阵,并结合层次聚类算法实现户变关系的识别。最后,实际算例结果表明,提出的方法相比于现有方法准确率更高,误报更少,能直接应对数据缺失的情况,且具有更高的效率。

In view of the possible problem of wrong user-transformer relationships in the topology file of the low-voltage distribution station area,this paper proposes a user-transformer relationship identification method based on multi-feature symbolic aggregate approximation(MF-SAX)and hierarchical clustering.First,the symbolic aggregate approximation expression method is used to convert the user voltage time series into string series,and two additional parameters,i.e.,the voltage fluctuation coefficient and the voltage change trend,are introduced to strengthen its feature expression.Then,the similarity matrix of the user voltage curves is generated based on the edit distance,and the hierarchical clustering algorithm is combined to realize the identification of the user-transformer relationships.Finally,the results of the practical case show that,compared with some existing methods,the proposed method achieves higher accuracy and fewer false alarms,which can directly respond to missing data situations,and has higher efficiency.