This paper presents a two-stage stochastic programming formulation for yearly generation maintenance scheduling when the power generators are unreliable, which means they may be not available to provide service. Each generation company owns several power units and makes its own unit’s maintenance decisions to maximize the net profit, considering the failure of the units is probable. At the first stage, maintenance schedule decisions are set. Once the maintenance schedule has been fixed, generation amounts and load allocations are determined in the second stage. The second-stage decisions are made in response to the realized scenario. Since generation companies’ decisions are independent, they are considered as a Nash Equilibrium. After the Nash Equilibrium obtained, reliability indexes are acquired by the independent system operator in order to prepare rescheduling signals. The capabilities of the proposed model are assessed and evaluated on a modified IEEE reliability test system.