This paper deals with the study of stochastic parameters of external loads, which are used in the tasks of determining the level of reliability of steel storage capacities. To describe the random load process, the normal law and the double exponential distribution of Gumbel were used. It was formulated the technique of transition from the study of the entire random process to the consideration of its maximums. It was obtained quantitative values of the stochastic characteristics of snow and wind loads on the territory of Ukraine, without reference to the zoning maps. A general procedure for determining the probability of failure was formulated, depending from a given characteristic maximum. This index corresponds to the basic level of load. The average intersection of this level by a random load process is equal to one. Analytical formulas are obtained to determine the scale and position of the double exponential distribution of Gumbel, which depend on the characteristic maximum, as well as formulas, which help to calculate the statistical characteristics (standard, expected value and coefficient of variation) of the random value of the load maximums. The possibility of using this approach is theoretically confirmed when the density distribution of the ordinate of a random process follows the normal law. It was proposed expressions for the parameters of the distribution maximums of the random processes, which are described by the polynomial exponent and the Weibull law.