Substation distribution transformers require special attention for their high costs, long replacement times and for impacting, when they fail, directly on the reliability of the electric network, thus causing the interruption of several costumers. In general, substations are designed with transformers in parallel, so that the peak load can be met even if a transformer in the group is damaged. However, this solution requires high investments. An alternative is to share stocks of spare transformers and mobile substations. Stock sizing should be done with caution, as a large number of equipment can unnecessarily increase investment costs, and insufficient stock can decrease system reliability and increase operating costs. This dissertation presents two methods for stock dimensioning based on Markov modeling and on Monte Carlo simulation, which are compared in terms of their results and characteristics when applied to the same system. Both models were used to scale the stock for catastrophic and repairable failures, separately. The probabilistic method based on the Monte Carlo simulation was applied to a group of 115 kV and 15 MVA class transformers, aiming to demonstrate the method s ability to find solutions that are technically and economically compatible.