Abstract: By analyzing the time varying characteristics in the conditional skewness and kurtosis of wind power time series,the generalized auto-regressive conditional heteroskedasticity with skewness and kurtosis（GARCHSK） model was proposed as a novel wind power forecasting method. To test the time varying characteristics in the conditional moments of wind power time series, an efficient test, chain test, was proposed. By combining the parameter estimate of GARCHSK model, the practical processing method for parameter constraint was presented and the estimate efficiency was improved. Based on the historical practical wind power data of wind farms in Jiangsu province, the proposed GARCHSK wind power forecasting model was carried out and the time-varying conditional moments of wind power time series were analyzed. Employing the Quasi Conditional Maximum Likelihood Estimation（QCMLE） based on Gram-Charlier series, the parameters of GARCHSK were estimated. The forecasting result of GARCHSK wind power forecasting model shows the feasibility and effectiveness of the proposed method.