SUMMARYAnalysis of electricity consumption behaviors lays the foundation for power grid planning, demand‐side response, electricity pricing, and energy efficiency improvement. In this study, we used the principal component analysis (PCA) to reduce the dimensionality of electricity load data and used the density peak clustering algorithm based on K‐nearest neighbors and shared nearest neighbor similarity (DPC‐KS) for cluster analysis of load profiles so as to obtain the electricity consumption behaviors of customers. In DPC‐KS, the local density was defined by integrating the idea of K‐nearest neighbors to find the density peaks, which promoted the accuracy of the cluster centers found. Also, a sample similarity measure criterion of shared nearest neighbor similarity was defined, a sample similarity matrix was built, and the samples were allocated accordingly, which enables a more accurate allocation of the remaining samples. Additionally, PCA and DPC‐KS were used to conduct cluster analysis for the electricity load data of 315 dedicated substation customers in a region, and four types of electricity consumption behaviors were obtained and analyzed. The experiments validated the effectiveness of DPC‐KS, and DPC‐KS provided technical support for intelligent decision‐making in the power grid.