Abstract: With the continuous advancement of the country's "dual carbon" goal, the proportion of installed wind power generation continues to increase, and the large-scale wind power output with strong random fluctuations poses severe challenges to the "guaranteeing consumption and power supply" of the power system. High-precision wind power prediction is an important basic means to solve the above challenges, and in wind power plants and grid dispatching centers, the improvement in the accuracy of wind power prediction has continuously been considered as a long-term priority. Consequently, a wind power prediction correction method based on statistical analysis of short-term wind power prediction error distribution characteristics and fluctuation characteristics is proposed. Firstly, the error temporal-conditional characteristics are taken into consideration, and an error probability density distribution analysis based on improved nonparametric kernel density estimation (KDE) is conducted to obtain confidence intervals for wind power prediction at different confidence levels, in order to achieve hierarchical division of prediction errors. Secondly, by using variational mode decomposition (VMD) algorithm, the wind power prediction error sequence is decomposed into trend components and random components, and classification prediction is carried out according to the characteristics of the two types of error components. Finally, the hierarchical division results of errors are combined with the analysis of error fluctuation characteristics, thus a comprehensive judgment is made, and error compensation schemes for various situations are proposed to obtain corrected short-term wind power prediction values. The actual example shows that the proposed error compensation method can be adopted to reduce the monthly root mean square error of wind power generation by 2.6 percentage points compared to that before compensation, and the average absolute error by 2.4 percentage points compared to that before compensation. This method can be adopted to effectively reduce wind power prediction errors and improve short-term wind power prediction accuracy.