One of the first steps in distribution networks development planning is to forecast the loads, loads that the networks will supply. The load estimation used in electrical networks planning must consider not only the future loads, but also their geographical positions, for permitting the designer to locate and to dimension the electrical equipment. The load estimation influences various aspects of distribution system planning such as peak load demand period, transformer sizing, conductor sizing, capacitor placement, and so on. Hence, the peak load estimation is a problem of a special interest. The most efficient method in estimation of the peak load consists in utilization of the periodic energy consumption by customers who are divided into classes having different load shapes. In the first section of this work an analysis of the estimation methods for the peak load is presented. In the next sections a new method, based on the nonlinear fuzzy regression is described. The choosing of the method depends on available data, the imposed goal and accuracy.