With the growth of China’s economic strength and the continuous adjustment of the economic structure, the share of electricity consumption of residents in the whole society is gradually increasing and there is a trend of continuous increase. Appropriate and reasonable prediction of the level of electricity consumption for residents will be for power planning and Demand side management provides the basis for decision making. In this paper, the adaptive fuzzy C-means algorithm is used to cluster the resident workday load data and holiday load data. According to the clustering results, the load characteristics analysis is carried out, and a new user classification method is proposed to realize the differential analysis of the residential electricity behavior characteristics. This classification method can not only accurately describe the power behavior characteristics of power users, but also provide effective data support for demand response such as peak hour price, peak balance management, and load regulation.