With the eventually improvement of power consumption information collection, the accurate analysis for user power consumption mode will provide an important basis for power intelligent construction. When analyzing power consumption modes, the load is the only clustering feature to be taken into account. Therefore, a weighted clustering analysis method considering multi-type feature selection is proposed. Firstly, the load and meteorological features are normalized to establish a feature set to be selected. Then, the clustering feature set is selected by combining mutual information and grey correlation degree. Finally, the weighted k-means algorithm is utilized to cluster the selected feature sets, and the typical behavior of each power consumption mode is analyzed with the load curve. Through the analysis of the electrical load data of a shopping mall in Shanghai, it is proved that this method can reduce the interference of redundancy information and improve the clustering quality.