User Seasonal Electricity Behavior Portrait Based on Data Feature Extraction and K-Means Clustering

Abstract

The safe and stable operation of China's AC/DC power grid and the consumption of new energy put forward high requirements for demand side response, and the image of power users is of great significance for the implementation of demand side response. This paper proposes a method of power user's electricity behavior portrait considering load seasonal characteristics based on data mining technology. Firstly, data cleaning and preprocessing are carried out on the original load data, and feature filtering is performed using variance filtering and feature filtering; then, according to the load characteristics, the processed load data is decomposed into base load and sensitive load affected by other factors; secondly, the correlation coefficients of the base load and the sensitive load are analyzed by sequential clustering, and the results of double clustering labels are obtained, and the final results are obtained by the two labels forming the Finally, the two types of labels are obtained to form a portrait of the user's electricity consumption behavior. The feasibility of the proposed user profiling scheme is verified in the case analysis section by using the load data of 30 power users.