With the widespread deployment of end devices, massive amounts of power data are generated. Aiming at the load data of power users, this paper proposes a Fuzzy C-Means (FCM) clustering algorithm based on Pearson’s correlation coefficient. Then we consider methods that combine cluster analysis and feature representation methods. We further propose a load classification method based on Pearson-FCM and piecewise aggregation approximation (PAA). Firstly, the data are divided into training set and test set, and the typical daily load profile of each user is obtained based on the training set data, which is used as the benchmark model for load classification. Secondly, for the test set data, the same method was used to obtain the typical daily load curves. Then, we classify them into the user type to which the curve model with the smallest distance belongs. The user type with the highest number of classifications is identified as the final classification result. Finally, the validity and accuracy of the proposed methodology is further verified through an art case study.