| As a kind of renewable and clean energy, wind power can make up for improving the present situation of fossil energy exhaustion, and has double efficacy of adjusting the energy structure and reducing environmental pollution, it will be the most promising green energy in21st century. While influenced by wind speed change characteristics, wind power has a lot of randomness, intermittent, uncontrollability and reverse load characteristics, which constitutes a significant pressure to power grid scheduling and power supply management.For achieving stable operation of wind power supply system, there must be adequate spare power source and load capacity, which will increase the power system operation cost largely, and the introduction of large-scale wind power has already brought a lot of hidden trouble to power system. With the increase of the installed capacity of wind power and wind power penetration power exceeds a certain value, in order to make the power grid operate smoothly, the effective and reliable wind power prediction system will become an important part of electric power system.Wind speed forecasting is the most important foundation and the premise of wind power and wind farm generated energy prediction, now wind speed forecasting model accuracy in China is still at unsatisfactory level, the wind speed prediction accuracy has become the existing problems to be solved presently.This paper has analyzed the wind change general rule of He’xi area in the past60years systematically, whose wind energy reserves is biggest in the Gansu province, divided wind speed change characteristics types, and put forward a new method of wind speed forecasting-wavelet decomposition autoregressive combined with wind speed change parting forecast method (WDAR2) by bring wind speed change characteristics types into wind speed prediction, which is suitable for the region, and made the foundation to improve the wind speed prediction and wind power evaluation accuracy, provided technical support for stable wind power deployment accessing to power grid.1. It is inconspicuous for the analysis of wind speed variation characteristics and results of the impact to change wind observation method; Before2000, the wind speed overall declines in He’xi area, while it is on the trend of growth after2000; The suitability is poor by using Mann- Kendall (M-K) test method to analyse wind climate change in He’xi area. In the years2005-2010, the maximum wind speed of Guazhou (formerly:An’xi) county area is18.4m/s, not more than25.0m/s.2. The wind speed change in He’xi area can be divided into four types:gentle wave type, common wave type (including peak type and broad peak type), increasing type and decreasing type.3. There are differences between wind power resources calculated from the observation data based on meteorological station above the ground10m level, wind tower above the ground70m level, and wind power farm above the ground70m level in the same area. It is needed to make wind correlation analysis and do the necessary corrections when using alternatives material such as the observation data of meteorological station or regional wind tower to calculate wind power resources of wind farms. At the same time, it is needed to correct the ground wind of meteorological station to fan height.4. The wavelet decomposition autoregressive combined with wind speed change parting forecast method (WDAR2) can be used to forecast wind speed in short-term, and the wind speed prediction accuracy is improved in a certain extent.5. During the short-term wind speed prediction methods that meet the power system control requirements, wavelet decomposition autoregressive combined with wind speed change points type of wind speed forecasting method (WDAR2) is better than wavelet decomposition autoregressive wind speed forecasting method (WDAR1), which is also superior to the current more widely applicable autoregressive wind speed forecasting method (AR). |