This work shows a novel method to characterize the power performance of a wind turbine directly from fluctuating data. In particular, it shows how to evaluate the dynamic responses of the power performance system on fluctuating wind speeds in the range of seconds. The approach is based on the stochastic differential equations known as the Langevin equations of diffusive Markov processes, which decomposes the fluctuating power production of wind turbines into their deterministic relaxation and stochastic-force (noise) contains. The operated steady-states of the complex power production system of a wind turbine on turbulent winds can be now properly reconstructed. The method is universal and not only more accurate than the current standard procedure but it also allows a faster and robustestimation of wind turbine power curves.In addition, the stochastic power output of a wind turbine is analyzed in the response theory.