Smart meters can collect power consumption data of consumers in real time, and will be widely popularized in the future digital power distribution system. In the context of the further liberalization of domestic electricity sales market and its gradual prosperity, electricity retailers can analyze the mass electricity consumption data of the consumers to grasp the electricity consumption behavior, thereby achieving better services. The feature extraction method for user behavior analysis of electricity sales market is discussed. K- means, fuzzy clustering, hierarchical clustering and other clustering algorithms are adopted to achieve the pattern extraction of typical user electricity behavior, and the basic features of electricity consumption behavior of different users are analyzed.An empirical analysis is conducted on the public electricity data of 6 445 consumers in Ireland. The result proves that the proposed method can effectively extract the patterns of behaviors, and distinguish the differences and similarities among users.