The error test of gust forecast has a certain guiding significance for the refined forecast correction in practice, and provides a reference for how to eliminate the influence of the daily variation of error in the refined forecast. The 10 m gust and 10 m average wind forecast data of the European Centre for Medium-Range Weather Forecast (ECMWF) are selected from the fine grid for 3-72 h day by day from 2017 to 2019, and based on the real maximum wind data 3 hours by 3 hours of 9 national meteorological observation stations in Dalian area, the error test of forecast is analyzed. The results are as follows: According to the forecast error test based on the forecast and the actual situation, the mean error（ME）of the ECMWF fine grid forecast is 0.96 m·s-1, which indicates that the forecast is larger on the whole. However, the statistical conclusions of the forecast errors of the two classifications are inconsistent for each wind level, and the test according to the forecast wind level is more consistent with the actual forecast work based on the model forecast. According to the statistical test of the forecast, the forecast errors of each wind direction, each wind level and each station are obviously different. The larger the wind level is, the greater the degree of forecast bias is, and wind direction also shows the trend of error increasing with

wind level increase. The average error of gust forecast has obvious daily variation, with the largest error around 08:00 and the smallest error around 20:00, which is mainly caused by the daily variation of the average error of 10 m average wind prediction. The correlation coefficients between all forecast cases and the observations for each predictive aging are above 0.7, and when it comes to each wind level and wind direction, the correlation of each wind direction is good, but the correlation of each wind level is significantly reduced, and the reliability of the wind forecast of the magnitude 8 and above is decreased greatly.

阵风的预报误差检验对实际工作中的精细化预报订正具有一定的指导意义，同时对精细化预

报中如何消除误差日变化的影响提供了借鉴。选取2017—2019年3~72 h逐日逐3 h 欧洲中期天气预

报中心（European Centre for Medium-Range Weather Forecast，ECMWF）细网格10 m阵风和10 m平均风

预报资料，基于大连地区9个国家气象观测站实况逐3 h极大风资料进行预报误差检验分析。结果表

明：按预报风级和实况风级分类的预报误差对比检验均表明ECMWF细网格预报整体偏大，平均误差

为0.96 m·s-1，但具体到各风级时两种分类的预报误差统计结论并不一致，按预报风级分类的检验更

符合基于模式预报开展的实际预报工作。以预报为基准统计，各风向、各风级、各站的预报误差均差

异明显，风级越大预报偏大的程度越高，风向也表现出随风级增大误差增大的趋势。阵风预报的平均

误差具有明显日变化，08：00（北京时，下同）前后误差最大，20：00前后误差最小，主要由10 m平均风

的平均误差日变化所致。全部预报个例与实况各时效预报相关系数均在0.7以上，具体到各风级、风

向时，各风向相关性均较好，而各风级的相关系数则明显降低，8级及以上风力预报的可信度大幅

下降。