For planning the operation of power transmission systems, which transport the energy produced by generation plants to customers centers, it is essential to establish when and which transmission lines are unavailable due to preventive maintenance. In a previous work, we proposed a mixed-integer formulation for the transmission maintenance scheduling problem that schedules preventive maintenance while keeping network connected and maximizing customers' supply in case of unexpected line failures. Starting from this previous deterministic model, we now propose a mixed-integer formulation that considers the uncertainty on preventive maintenance duration, over a one-year period. This stochastic formulation considers the existence of an unknown delay in maintenance tasks at the planning time. Task delays are represented by scenarios with associated probability of occurrence. This large-scale problem is solved using a proposed algorithm which decomposes the formulation into a master problem, solved with CPLEX through Benders decomposition, and sub problems that validate the solution found. We present computational results for the 24-bus IEEE network that demonstrate that the proposed algorithm reaches optimality more efficiently than the direct solution of the stochastic extensive formulation. When the algorithm runs larger instances the optimality level achieved is not as good as for smaller instances, but it presents a very good tolerance gap equivalent to the default CPLEX/AMPL value.