Within the framework of the task of assessing the balance reliability of electric power systems (EES), the calculation procedures for determining probabilistic indices of power deficit in concentrated EES are considered. It is shown that the best method for estimating the desired indicators is to separately account for generation and load. The place of complete convolution of probabilistic power generating series is shown when estimating the error of probabilistic modeling of the states of generating groups. For comparison and justification of analytical models, the number of tests in the statistical modeling method (Monte Carlo) that is admissible to ensure acceptable accuracy of the solution has been determined. The admissibility and the possibility of practical application of the normal distribution to describe the generation system when determining the probability parameters of the power deficit in a concentrated EPS with a normally distributed load power is considered. A mathematical procedure is proposed for correcting the distribution function of the unbalance of power by the Gram-Charlier expansion, which allows us to significantly improve the quality of the solution compared to the application for the system of generating the normal distribution without correction. The possibility of using von Mises method for combining different types of generator groups with different probabilistic distributions of available power is stated.