Research of pattern recognition on resident power consumption behavior based on massive data plays a crucial role in power grid load prediction and demand response potential development. However, the emergence of large-scale data and multi-dimensional features of residents’electricity consumption has become a difficulty in identifying the heterogeneous pattern of residents' electricity consumption. Firstly, based on the large-scale residential electricity consumption data, the power load characteristic decomposition technology is used to construct the feature project. And then, the multi -dimensional features constructed are fused by factor analysis, and the clustering algorithm is adopted to identify the residential electricity consumption pattern. Finally, taking Jiangxi province residents’electricity consumption data as an example, an empirical analysis is made in both households and community level, and four typical residents’electricity consumption patterns have been obtained, which can provide scientific support for the grid companies to make tailored and differentiated policies and further expand the depth and breadth of services.