By using the hourly extreme wind from the conventional meteorological observation and the 6 h extreme wind forecast data from ECMWF between February 2015 and February 2018, the representative stations in the Bohai Sea were selected to analyze the prediction error characteristics of the extreme wind products in the ensemble forecast. The analysis showed that the spread of the extreme wind product in ensemble forecasting is obviously smaller than the root mean square error, and whether the forecasting results of each forecasting member are concentrated or not cannot reflect the reliability of forecasting. Due to the limitation of model prediction ability, it is impossible to select the most credible prediction results simply by ensemble prediction. The ensemble average, the 75th percentile and maximum ensemble forecasting have their own advantages and disadvantages in extreme wind forecasting. Therefore, based on the above three statistics and the frequency of different magnitude wind speed occurrence, an objective correction method for extreme wind forecasting in the Bohai Sea is established. The comparative analysis of experiments shows that the correction method can effectively improve the accuracy of extreme wind forecasting, which can provide important references for the forecast of extreme wind process.