Under the background of massive heterogeneous flexible resources participating in the operation regulation of power grid with high proportion renewable energy, an analysis method for power load clustering based on complete feature index and improved density peak algorithm to the high requirements of accuracy, robustness and computational efficiency in the analysis of user power consumption characteristics is proposed in this paper. Firstly, nine complete characteristic indexes are extracted for index reduction and improvement to replace the power vector composed of daily load curves as clustering input. Secondly, the entropy weight method is used to assign weight to each characteristic index to ensure the morphological characteristics of load curves. Finally, an improved density peak clustering algorithm is applied to clustering analysis of daily load. Case studies are carried out based on a certain area actual load data, and the results show that the proposed method is superior to the traditional power load clustering algorithm in terms of robustness and clustering quality. The clustering results can reflect the actual power consumption characteristics of users truly and effectively, which will provide a situational awareness basis for formulating accurate user portrait and demand response strategy.