The purpose of the thesis is to test alternative methods for estimating the extreme response for wind turbines under normal operation. Currently, different extrapolation methods are used for determining the extreme responses by extracting maxima from a few simulations of normally 10 minutes. These extrapolation methods are associated with unknown statistical uncertainties. In the first part of the thesis, the extrapolation method suggested by the IEC-61400 standard is described and the associated uncertainties and problems are clarified and the extrapolation method is tested for determining the extreme response for a wind turbine model. Afterwards two simple structural systems are presented, a simple shear frame and a simple wind turbine model, which will be the basis for evaluation of the alternative methods presented. The alternative methods Russian Roulette and Splitting with Distance Control (RR&S), Importance Sampling (IS) and RESTART are presented as alternatives to the present extrapolation methods. The latter of the three does not provide promising results, whereas RR&S and IS provide promising results on both of the simple structural systems reducing the computational time significantly. Finally, the RR&S algorithm is implemented on the aero-elastic code FAST. The algorithm provides promising results for estimating the extreme responses if disabling the wind turbine control system and no significant improvements if the control system is enabled.