This paper proposes a load estimation algorithm based on K-means clustering analysis. In this algorithm, the clustering load curve and distance function are applied to estimate the missing and future measured values. The correlation distance functions of Canberra, Manhattan, Euclidean and Pearson are studied. Some case studies are carried out by using the daily and segmented load curves of the aggregation smart meter. When the time window covered by the segmented configuration curve is less than or equal to 24 hours, the simulation results show that the Canberra distance is better than other distance functions; the results also show that the segmented cluster center produces more accurate load estimation than the daily cluster center, and the cluster center can obtain more accurate estimation within the range of 16 – 24 hours. According to the research results of this paper, the developed load estimation algorithm can be integrated with state estimation or other network operation tools, which can better monitor and control the distribution network and provide services for the power system.