Analysis of residential electricity consumption behavior based on WTK-medoids clustering

Aiming at the problems of data redundancy and complex clustering caused by the high dimension and noise of residential user load data in the process of data analysis, a residential electricity consumption behavior analysis method based on WTK-medoids clustering is constructed. Firstly, an improved wavelet threshold fonction for noise removal is proposed to address the characteristics of noisy residential user load data, and secondly, the noise reduced load data is used as input to Principal Component Analysis (PCA) to reduce the dimensionality of the data and extract the principal component characteristics. Finally, the load data processed by filtering and dimensionality reduction is used as the data source, and the K-medoids algorithm is used for clustering. Compared with the single clustering algorithm, the proposed residential electricity consumption behavior analysis method based on K-medoids clustering with improved wavelet threshold has better user segmentation effect.