The lifetime of transmission and distribution power systems is long and thus, long-term plans are needed for their successful development. In generating long-term scenarios, the starting point is the analysis of the present electricity consumption. The data of electricity consumption will become more exact by the end of 2013, when hourly based automated meter reading (AMR) consumption data will be received from each customer in Finland. The amount of data is huge and powerful analysis methods are needed. This paper presents a method for clustering the electricity consumptions using principal component analysis (PCA) and K-means clustering. AMR data of 18 098 customers from two city districts of Helsinki, Finland is applied for a case study reported in this paper. A multiple regression analysis is also carried out on the two largest clusters to find the most important explanatory factors for the load modeling. The interpretations of the clusters and the plausibility of the regression coefficients are considered very important. Five distinct and meaningful clusters are found. The regression models give interesting insights into the explanatory factors behind electricity consumption. The models of the main customer groups assist the distribution system operator (DSO) in the long-term development of the power system.