This paper proposes a novel unit maintenance scheduling（UMS） model， considering the influence of unexpected unit failures. The uncertain factor not only leads to costly expenditure for replacing the damaged components， but also lowers the system reliability and increases the operation cost. This is rarely considered or is often simplified as a constant value in the existing approaches. Based on the generator's bath-curve failure model and the corresponding renewable costs， the impacts of scheduling outage and unexpected failures on the operational costs are discussed in detail. The reliability costs and generation costs of the system， and the renewable costs and maintenance costs of generating units are included in the operational costs analysed. The former two costs are calculated through the system probabilistic production simulation by equivalent energy function（EEF） method. Subsequently， the UMS model is presented to minimize the total costs over the whole planning horizon， which is solved by a genetic algorithm because of its non-linear and non-differentiable characteristics. Finally， numerical examples based on a 21-unit system are utilized to demonstrate the usefulness of the proposed scheme.