Abstract: Aiming at the deficiency of correlation-based wind speed generated by the existing correlation matrix method, a mixed half-cloud model of wind speed is introduced in this paper to improve the sampling process of wind speed. The Monte Carlo sampling is used to generate correlation-based wind speed samples. Comparison between historical wind speed and model results shows that the correlation coefficient between the wind speeds generated by the proposed improved correlation matrix method is more similar to the real data. Then, different correlation coefficients are selected to analyze the wind farm output and grid connection reliability, the results show that the correlation coefficient has great impact on the wind farm output and reliability. By considering the correlation of wind speed,the more accurate analysis on wind farm output and connection reliability can be obtained.