With the increasing penetration of wind power into power systems， wind power forecast becomes increasingly important. Non-parametric regression， one of typical model evaluation and prediction methods， has not been used in the domestic short-term wind power forecast. This paper applies non-parametric regression model to the short-term wind power forecast of point value and probability interval. First， the forecast model for the point wind power with the wind speed as the input variable is developed based on the non-parametric regression. Then， the probability distribution function of wind power forecast error is deduced from an empirical distribution model and non-parametric regression technique to analyse the probability interval of point forecast value. A wind farm in Inner Mongolia is chosen as an example to verify the effectiveness of the model in the short-term wind power forecast.