Abstract
ABSTRACT Wind speed variables play an important role in exploiting wind power. However, they are fluctuating and random. Therefore, understanding their characteristics and properties is necessary to improve exploitation efficiency. This research investigates various wind speed distribution models, both parametric and nonparametric, to estimate wind speed probability density (WSPD). The distribution models are implemented on various wind speed datasets with distribution characteristics of varying complexity. The assessment of goodness of fit includes statistical tests including Cramér‐von Mises (CvM), Anderson‐Darling (A‐D), Kolmogorov‐Smirnov (K‐S), and chi‐square (), along with indices correlated as mean absolute percent error (MAPE). The study highlights that the adaptive bandwidth kernel density estimation (AKDE) distribution model based on the nearest neighbor estimation (NNE) has superior goodness of fit performance. Wind turbine power curves are applied to calculate and compare expected, distribution‐based, and empirical power output. In addition, the difference between the power output estimated from the AKDE distribution and the estimate from the empirical wind speed is almost zero, so this estimated power is reliable and can be used as a reference for planning or evaluating wind farm efficiency.