Estimation of the power demand for a plant having machines whose individual demands fluctuate presents an important problem. This problem is of significance to an electric utility if the anticipated power demand is likely to prevail over the existing demand in the area. Statistical techniques which use data, based on past history of the demand for the area, are inadequate when the plant is under design. An analytic method to obtain probability distribution functions of the demand for the plant from information available at the design stage is presented in this paper. Typical power demands are classified and defined as random variables. Probability distribution functions of those variables are obtained for individual machines. These are then used to derive the distributions for the total demand of the plant. Utilization of the distributions is illustrated by examples.