An analysis method for residential electricity consumption behavior based on UMAP-CRITIC feature optimization and SSA-assisted clustering

Abstract

The analysis of residential electricity consumption behavior (RECB) aims to deeply reveal the electricity consumption characteristics of customers that can be used to improve the electricity service performances from the large volume of load data. For analyzing the RECB, various clustering-based methods have been developed. However, with the dimension of power consumption data increasing, it is difficult for traditional methods to accurately identify users’ power consumption patterns. To this end, this paper proposes a novel analysis method for RECB with high-dimensional data based on uniform manifold approximation and projection-criteria importance through intercriteria correlation (UMAP-CRITIC) feature optimization and sparrow search algorithm (SSA)-assisted clustering. Specifically, the high-dimensional raw data combined with load characteristics indexes are first utilized to generate low-dimensional features by UMAP-based dimension reduction theory. Secondly, the contribution weight of each reduced feature to the intrinsic information of the original data is optimized based on the CRITIC weight method, and then the data feature set of RECB identification considering feature contributions is constructed. Finally, the SSA-based k-means is employed to obtain and analyze the various RECBs. The simulation results show that the proposed method can effectively identify high-dimensional electricity consumption data and precisely extract the electricity consumption information of residents.

Keywords

• Customer electricity consumption behavior analysis
• Dimension reduction
• Load characteristic indexes
• Clustering
• SSA