A stochastic optimization approach is proposed for the unit commitment problem with the uncertainty of wind power generation taken into account， based on mixed-integer linear programming（MILP）. The problem is formulated to minimize the total operation cost of thermal units. In considering wind power generation， scenarios are generated by auto-regressive and moving average（ARMA） time series model and Latin hypercube sampling（LHS） method and the stochastic optimization problem is then transformed to a deterministic one. A large number of scenarios lead to computing complexity， scenario reduction technology is introduced to decrease scenario number in order to reduce computing cost. The proposed formulation is tested on a 10-unit system and a 100-unit system. Simulation results show that the varying wind power generally leads to the increase of total cost. In addition， the ramping rates of non-wind generators and the prediction precision of wind power are significant in making generation scheduling with volatile wind power generation. Moreover， the system operation cost decreases significantly if wind power is considered as a spinning reserve resource.