Abstract: When the power output of wind power exceeds the grids acceptable capacity, it is needed to limit the power output of wind farm. According to prediction information of wind farm, wind power limit is allocated initially according to the proportion. The random variables correlation analysis is used as the theoretical basis to describe the influence of power output correlation of wind farms on the fluctuation of overall wind power output. With the aim of reducing further wind power loss, the wind farms power limit is revised based on the correlation coefficient between power outputs of wind farms by considering the difference of installed capacity of wind farms. A practical wind farm group consisting of 5 wind farms located in a region is studied to testify the validity of power limit correction method. The result shows that the actual power output of wind farm is closer to the ideal power output curve, and abandoned power capacity is reduced, which verifies the effectiveness of the proposed method.