基于异常点检测和改进kNN算法的台户关系辨识方法研究

Research on the Method of Identification of the Relationship Between Transformer and User Based on Abnormal Point Detection and Improved kNN Algorithm

台户关系识别是电网公司实现营配贯通的基础。为此,提出一种基于异常点检测和改进kNN算法的台户关系辨识方法。首先,利用局部异常因子算法剔除不属于待辨识台区的用户,保证待辨识台区和用户具备映射关系。然后,对传统的kNN算法进行改进,避免k值选取带来的过拟合或训练误差较大等缺陷,提升算法的准确性和鲁棒性,最终实现台区用户的有效辨识。算例分析表明,所提的台区用户辨识方法能够准确有效识别台户关系,且更适合实际辨识工况。

Identification of the relationship between transformer and user is the basis on which the power grid company realizes the connection between marketing business and production business.For this reason,a method based on abnormal point detection and improved kNN algorithm was proposed in this paper for identification of transformer-user relationship.Firstly,the local anomaly factor algorithm was used to reject those users not belonging to the transformer area to be identified,so as to ensure mapping relationship between the transformer area to be identified and the user.Then,the traditional kNN algorithm was improved to avoid over-fitting or large training error caused by k value selection,improve the accuracy and robustness of the algorithm,and finally realize effective transformer area and user identification.Example analysis showed that the proposed identification approach could identify the relationship between transformer and user and was quite suitable for actual identification conditions.