| Summary: | Precisely identifying the household-transformer relationship is of significant importance for both the stability of the power system and the quality of customer electricity consumption. However, the complex network structures and frequent reconfigurations may lead to inaccurate records of household-transformer relationships. In this paper, a novel data-driven similarity assessment solution is proposed to enhance the accuracy and scalability of identifying household-transformer relationships. Initially, a data processing method based on dynamic temporal regularization with sliding windows is employed to optimize dataset quality as well as enhance the efficiency of data processing. Then, a two-stage solution is proposed for identifying the household-transformer relationship. The first stage involves initial normalized clustering based on the basic information of power distribution substations, while the second stage assesses the similarity between households and transformer operational states based on Hausdorff distance. The superior performance of the proposed method is extensively assessed through real historical datasets, compared to benchmarks.
|
|---|