Detection of abnormal power consumption behavior including energy theft is an important and difficult part of electricity inspection. Due to data aggregation problems, previous studies have mostly focused on electricity stealing behaviors of large customers, and research on residents'' electricity stealing behaviors is relatively weak. This article extracts electricity characteristics from data collected by smart meters, and then uses fuzzy clustering to analyze the data structure to extract behavior characteristics of normal users. Experimental results on real-world dataset show the applicability of the proposed method.
