One of the most important needs in planning and analyzing the operating modes of the power system is forecasting the quantities and load schedules. The paper suggests a search algorithm for the same type of load graphs based on data mining methods. The initial data is formed by the averaged half-hourly graphs of consumption of active and reactive power. The proposed algorithm is primarily aimed at preliminary processing and grouping of measurement data when solving the task of identifying static voltage load characteristics, but it is also possible to carry out research in the area of searching for typical load schedules for design and prospective calculations. The result of the data processing will be an array of labels to link the daily chart with a particular group. The proposed algorithm is based on the k-means method. The algorithm is tested on real data collected from the substation feeding the assembly plant. If the initial data is the sampling of the load curves for several years, the task of predicting the configuration of the load graph can be set. The initial data for solving this problem will be an ordered array of timestamps for the assignment of the daily load graph to a particular type.