In a bid to detect abnormal electricity consumption of power users, a method for identifying abnormal electricity consumption mode based on unsupervised combination algorithm is proposed in this paper. The proposed identification method consists of three parts of data processing, feature extraction and outlier detection. The power consumption and related data of the users are obtained, and the data is cleaned and the missing value is supplemented, feature extraction is carried out on the data to obtain the corresponding features for abnormal electricity consumption recognition. Afterwards, k-means is used to cluster the users into two groups, and principal component analysis is performed on each group to optimize the feature space, the outlier proximity is calculated, and abnormal power consumption users are identified by 2 sigma principle. This method improves the identification efficiency by combining clustering, optimization of feature space and outlier detection. The simulation experiment of abnormal power consumption user identification is carried out with real power consumption data, and the identification results verify the effectiveness of the proposed method.