In wind energy conversion systems, a power curve links the wind speed to the power produced by a wind turbine and an accurate power curve model helps wind power providers to capture the performance of wind turbines. For this purpose, this paper presents the implementation of novel hybrid approaches to the power curve modeling process of wind turbines. As a result of employing the complementary phases called clustering, filtering and modeling in this process, the k-means-based Smoothing Spline hybrid model achieves the most accurate power curve in terms of sum of squared errors, coefficient of determination and root mean squared error. On the other hand, the k-medoids + + -based Gaussian hybrid model causes the most inconsistent power curve in terms of the mentioned goodness-of-fit statistics. Furthermore, all of hybrid power curve models constructed in this paper outperform the conventional linear, quadratic, cubic, exponential and logarithmic benchmark models with the high improvement percentages. Finally, the proposed hybrid power curve models are shown not to be dependent on the initial raw power curve data.