中文摘要：

应用MM5模式以及1°×1°的NCEP再分析资料，对新疆风能资源“代表年”进行了3km×3km分辨率的模拟试验，同时将10In高度的模拟值与气象站10m高度的实测值进行了对比。结果表明：（1）模式能较真实反映年、月平均风速大小的空间分布特征，但存在一定的系统性偏差，且偏差的大小具有明显的月际变化特征与地域性变化特征。平均来说，夏半年的偏差幅度小于冬半年，大风区的偏差幅度明显小于非大风区，达坂城一小草湖风区、哈密东南部风区、三塘湖一淖毛湖风区的模拟偏差最小。（2）对各风区逐小时平均风速模拟值进行线性回归订正，虽能有效减小有些风区模拟与实测值间的风速偏差，但对有效风速小时数的订正效果极其有限，订正后的偏差仍具有随机性。（3）以月为单位，通过对逐小时平均风速模拟值立方的回归订正可有效减小年平均风功率密度的模拟误差，同时模式中逐小时的空气密度可直接以观测点的月平均空气密度取而代之。该试验不仅对近期新疆已经完成的风能资源详查与综合评价中有关中尺度模式参数化方案的最优组合选择和水平分辨率的调整具有现实意义，而且也为如何提高风电功率短期预报的精准性提供了研究素材。

英文摘要：

When the mid-term plan about wind power development and reaching interconnection objective between Xinjiang and Northwestern of China Power Networks in 2010 were issued by Xinjiang Uyghur Autonomous Region in 2008, the national wide detail wind energy resources investigation and estimation was still in the process of preparing the technical scheme and selecting sites of building wind measurement masts in the detail investigated areas. Under the circumstances, wind energy resources numerical simulation experiment of horizontal 3 km resolution was carried adopting mesoscale model MM5 in most of areas in Xinjiang based on the NCEP re-analysis materials in 2006, in which ,the wind energy resources characteristic has multi-year average representation to some extends, the analyses of comparison between the observed wind datum from the meteorological stations and the simulated ones at the height of 10 m above the ground indicate that the model has better ability to depict spatial distribution characteris- tics of either annual or monthly mean wind speed though the systematic bias exists to some extends. The bias ampli- tude varies with seasons and different wind areas, generally showing less bias in summer than in winter and less in strong wind areas than in weak wind areas. For example, the bias is the smallest in Dabancheng - Xiaocaohu, the southeastern of Hami, Santanghu - Naomaohu wind area. In some wind areas, the bias between the simulated and the observed annual mean wind speed can drop largely by the method of linear regressive modification month by month, but it is extremely limit for annual effective hour because the bias amplitude of modified simulation value is still random. The simulation bias of annual mean wind power density can drop largely with reducing hourly mean wind speed＇s cube by the approach of linear regressive modification month-by-month, while the air density can be replaced by the observed for a simulated month. The study doesn＇t only have practical significance for the recent finished de