A Method for Detecting Outliers and Identifying Typical Power Consumption Patterns in Low-Voltage Station Area Measurement Data Based on Two-Step Cluster Analysis

This paper proposes a method for detecting outliers and identifying typical power consumption patterns in low-voltage station area measurement data based on two-step cluster analysis. First, considering that high-dimensional daily power load sequences will lead to a decrease in clustering effect, five features are defined to characterize high-dimensional daily power load sequences. Secondly, the Density-based spatial clustering application with noise (DBSCAN) method is used to detect outliers in electricity consumption data. DBSCAN divides clusters based on the density of data points, and can effectively handle data sets with uneven density. It can effectively distinguish between points in low-density areas and points in high-density areas, which can completely overcome the inapplicability of traditional outlier detection methods. Disadvantages of changing data distribution. Finally, the power consumption pattern of typical users is extracted through the low-voltage station area user characteristic analysis method based on K-means clustering, and typical power consumption pattern analysis is carried out. This paper also takes the power consumption data of a certain low-voltage station area as an example to verify that the method proposed in this paper can better handle complex power consumption data, and can also effectively detect outliers and identify typical power consumption patterns, providing a System management and optimization provide a new idea and method.