Abstract: Large-scale wind curtailment results in several power-limiting outliers, which significantly affect the voltage problems caused by the grid connection, frequency modulation, and stability of the power system. In this study, the grey wolf-optimized boxplot algorithm（GWO-boxplot） is optimized for the adaptive identification and fine division of power-limiting and discrete outliers of each wind turbine in a wind farm. Considering the existence of a certain correlation between the output power of wind turbines in time and space, a wind turbine outlier processing method that integrates spatial and temporal correlations and weight coefficient distribution is proposed. The weight coefficients of the historical power-filling data and the spatial-correlation filling data were assigned, and a weighted summation was performed for outlier processing. Finally, the curves were fitted using a fractal interpolation algorithm. The proposed method was verified using the actual output data of each wind turbine in a cluster wind farm. Simulation results show that the proposed method can significantly improve the recognition and fitting accuracy of outliers.