Abstract: Integrating wind power that contains uncertainty and volatility in a large scale would bring great challenges to the operation and scheduling of power systems. The research on the unit commitment problem inside a wind farm contributes to improving operating efficiency of wind turbines,increasing output power of wind farm,and reducing the operation cost. Firstly,historical data of wind turbines is analyzed,based on which a comprehensive modeling method is proposed. Then,through the method dynamic probability distribution model can be built,based on which the characteristic matrix of wind turbine that changes over time is formulated. Secondly,the fuzzy C-means（FCM） clustering algorithm is used to classify wind turbines within a wind farm dynamically. Considering classification,the turbines are further labeled as conventional and dispatchable in different periods. At last,optimal scheduling model of a wind farm is proposed with the objective of minimizing fatigue damage of wind turbines. The effeteness of the proposed model is verified by the case study.