In a large-scale wind farm， interactions among the running wind turbines will affect the wind turbine power output and yield the wind energy loss. The wind energy loss directly affects the annual power productions， generation schedule and wind farm operation control. This paper presents a novel iterative regression method to eliminate the observation points with large errors under the abnormal conditions， so as to estimate the power output curve and calculate the wind energy loss for the operated large-scale wind farm. Finally， the actual data of a wind farm are used for calculating and analyzing the wind energy loss.