利用ECMWF集合风速预报产品和目前运行的风能专业数值模式预报产品，通过对比检验分析，设计一种适用于风功率短期预测、电网调度所需的风电场机组轮毂高度风速融合预报产品。即通过多统计量融合，分析其集合、最大最小及时空分布，通过概率分布函数、集合技术，研究提取波动风速的概率信息，指示出最有可能的风速波动位置，对集合存在的量级偏差进行概率匹配技术校准。尝试对各成员贝叶斯概率预报进行融合，获得代表ECMWF集合预报不确定性的集成贝叶斯概率预报，计算不同的集合统计量，形成单值预报，通过ARIMA提升预报的时间分辨率，基于BMA与确定性数值模式预报进行融合，给出机组轮毂最优风速预测曲线和最低最高可信区间。研究表明：融合预报对提升现有数值预报产品的精准度具有明显改进，验证期预报与现有确定性预报比较，平均绝对误差MAE降低了24.3%、相关系数R提升了12.5%；与ECMWF集合预报比较，MAE降低了11.7%、R提升了14.5%。
Using ECMWF (European Centre for Medium-Range Weather Forecasts) 100 meters wind speed ensemble forecast products and professional numerical model products on wind power, after the comparative test and analysis, designing a fusion forecast product of wind speed including the single value forecast and the wind speed variation zone in the studied wind farm, which is suitable for short term wind power prediction and electric dispatch. In the paper, the value of ensemble, maximum, minimum, and temporal and spatial distribution are analyzed according to multi statistic fusion, and then the probability information of extremely strong, weak and transition fluctuation of wind speed is researched and extracted using probability distribution function and ensemble technique, the most probable positions of the wind speeds fluctuation are pointed out, the magnitude deviation of the ensemble forecast products is calibrated by probability matching technique. The members of Bayesian probabilistic forecasting are attempted to fuse, thus the fused Bayesian probabilistic forecast representing the uncertainty of ECMWF ensemble forecast is obtained, the single value forecast is formed by calculating the different ensemble statistics at last. Using ARIMA(Autoregressive Integrated Moving Average Model) method to improve the time resolution of the forecast, and based on the BMA(Bayesian model averaging) method and fusion with the deterministic numerical model forecast, the optimal wind speed forecast curve and the wind speed variation zone is produced. The study shows that the BMA method is applied to the short-term wind power forecast and the dynamic fusion prediction model can reduce the forecast error effectively, the fusion forecast has obvious improvement on the accuracy than the ECMWF ensemble forecast and the existing numerical forecast products. Comparison between the validation period and the existing deterministic forecast, MAE decreased by 24. 3%, and the correlation coefficient (R) was increased 12. 5%. Compared with the ECMWF ensemble forecast, the MAE was reduced by 11. 7%, and the R was increased 14. 5%. In addition, the wind speed variation zone of the fusion forecast can capture the fluctuation of wind speed effectively and further reduce the risk on the electric dispatching plan and the wind farm operation decision caused by the single power prediction. The fusion of the existing numerical forecast products, is an effective way to improve the forecast accuracy of province whose computing resources are limited. Less number of study area selected and the study period is short in this study, more wind farms will be selected, and multi model ensemble forecast products will be fused next step. This method will be refined in order to promote and apply in all wind farms serviced.
[1]Baran S. 2014. Probabilistic wind speed forecasting using Bayesian model averaging with truncated normal components[J]. Computational Statistics & Data Analysis, 75(6):227-238.
[2]Sloughter M L, Gneiting T, Raftery A E. 2010. Probabilistic wind speed forecasting using ensembles and Bayesian Model averaging[J]. Journal of the American Statistical Association, 105(489):25-35.
[3]Chen Fajing, Jiao Meiyan, Chen Jing. 2011. A new scheme of calibration of ensemble forecast products based on bayesian processor of output and its study results for temperature prediction[J]. Meteor Mon, 37(1):14-20.<br/>陈法敬, 矫梅燕, 陈静. 2011.一种温度集合预报产品释用方法的初步研究[J].气象, 37(1):14-20.
[4]Chen Jing, Li Chuan, Feng Hanzhong, et al. 2004. Short-range ensemble simulation of "2003·08·28" heavy rain on the east side of Qinghai-Tibetan Plateau[J]. Plateau Meteor, 23(suppl):6-14.<br/>陈静, 李川, 冯汉中, 等. 2004.青藏高原东侧"2003. 8. 28"暴雨的集合预报试验[J].高原气象, 23(增刊):6-14.
[5]Fan Linjun. 2010. Preliminary study of statistically downscaled temperature ensemble predictions in eastern China[J]. Plateau Meteor, 29(2):392-402.<br/>范丽军. 2010.统计降尺度方法集合预估华东气温的初步研究[J].高原气象, 2010, 29(2):392-402.
[6]He Xiaofeng, Zhou Rongwei, Sun Yihan. 2014. Verification on surface wind speed of three global circulation models in China[J]. Plateau Meteor, 33(5):1315-1322. DOI:10. 7522/j. issn. 1000-0534. 2013. 00093.<br/>何晓凤, 周荣卫, 孙逸涵. 2014. 3个全球模式对近地层风场预报能力的对比检验[J].高原气象, 33(5):1315-1322.
[7]Jiang Ying, Song Lili, Cheng Xinghong. 2013. An integrated and revised method of forecasting wind speed for wind farms[J]. Resources Science, 35(3):673-680.<br/>江滢, 宋丽莉, 程兴宏. 2013.风电场风速预报集合订正方法的尝试性研究[J].资源科学, 35(3):673-680.
[8]Li Yan, Cheng Peipei, Lu Yixiong, et al. 2015. Wind power forecasting over the typical complex terrains[J]. Plateau Meteor, 34(2):413-425. DOI:10. 7522/j. issn. 1000-0534. 2013. 00181.<br/>李艳, 成培培, 路屹雄, 等. 2015.典型复杂地形风能预报的精细化研究[J].高原气象, 34(2):413-425.
[9]Liu Jianguo, Xie Zhenghui, Zhao Linna, et al. 2013. BMA probabilistic forecasting for the 24-h TIGGE multi-model ensemble forecasts of surface air temperature[J]. Chinese J Atmos Sci, 37(1):43-53.<br/>刘建国, 谢正辉, 赵琳娜, 等. 2013.基于TIGGE多模式集合的24小时气温BMA概率预报[J].大气科学, 37(1):43-53.
[10]Liu Jinda. 2000. Ensemble prediction ushered in new era in numerical weather prediction[J]. Meteorological Monthly, 26(6):21-25.<br/>刘金达. 2000.集合预报开创了业务数值天气预报的新纪元[J].气象, 26(6):21-25.
[11]Shi Chun'e, Li Yaosun, Yang Jun, et al. 2015. Comparison of simulations on winter sounding profiles in PBL in east China between WRF and MM5[J]. Plateau Meteor, 34(2):389-400. DOI:10. 7522/j. issn. 1000-0534. 2013. 00206.<br/>石春娥, 李耀孙, 杨军, 等. 2015.MM5和WRF对中国东部地区冬季边界层模拟效果比较[J].高原气象, 34(2):389-400.
[12]Shi Lan, Xu Lina, Hao Yuzhu, et al. 2015. Wind speed forecast products by BJ-RUC in a wind farm of Inner Mongolia[J]. Arid Land Geography, 38(3):510-516.<br/>石岚, 徐丽娜, 郝玉珠, 等. 2015.BJ-RUC在内蒙古某风电场的风速预报检验分析[J].干旱区地理, 38(3):510-516.
[13]Shi Lan, Xu Lina, Hao Yuzhu. 2016. The correction of forecast wind speed in a wind farm based on partitioning of The high correlation of wind speed[J]. J Appl Meteor Sci, 27(4):506-512.<br/>石岚, 徐丽娜, 郝玉珠. 2016.基于风速高相关分区的风电场风速预报订正[J].应用气象学报, 27(4):506-512.
[14]Song Lili, Zhou Rongwei, Yang Zhenbin, et al. 2012. Meteorological technology application and development in wind energy resources utilization[J]. Engineering Sciences, 14(9):96-101.<br/>宋丽莉, 周荣卫, 杨振斌, 等. 2012.风能资源开发利用的气象技术应用和发展[J].中国工程科学, 14(9):96-101.
[15]Sun Yue, Wu Tongwen, Jin Xia, et al. 2016. Evaluation of CMIP5 climate models on simulating the amount and frequency of convective and stratiform precipitation in East Asia[J]. Plateau Meteor, 35(6):1524-1539. DOI:10. 7522/j. issn. 1000-0534. 2015. 00107.<br/>孙悦, 吴统文, 金霞, 等. 2016.CMIP5气候模式对东亚地区对流和层云降水量和降水发生频率的模拟评估[J].高原气象, 35(6):1524-1539.
[16]Wang Li, Zhao Yuan, Yang Xianming, et al. 2015. Prediction of air quality in Lanzhou using time series model and residual control chart[J]. Plateau Meteor, 34(1):230-236. DOI:10. 7522/j. issn. 1000-0534. 2013. 00150.<br/>王莉, 赵渊, 杨显明, 等. 2015.基于时间序列模型与残差控制图的兰州市空气质量研究[J].高原气象, 34(1):230-236.
[17]Xiong Minquan. 2015. Climate regionalization and characteristics of surface winds over China in recent 30 years[J]. Plateau Meteor, 34(1):39-49. DOI:10. 7522/j. issn. 1000-0534. 2013. 00159.<br/>熊敏诠. 2015.近30年中国地面风速分区及气候特征[J].高原气象, 34(1):39-49.
[18]Zhang Dong'en. 1998. Estimation of parameter for truncated normal distribution via the EM algorithm[J]. Journal of Beijing Institute of Light Industry, 16(2):72-76.<br/>章栋恩. 1998.截断正态分布参数估计的EM算法[J].北京轻工业学院学报, 16(2):72-76.
[19]Zhao Linna, Liu Ying, Dang Haofei, et al. 2014. The progress on application of ensemble prediction to flood forecasting[J]. J Appl Meteor Sci, 25(6):641-653.<br/>赵琳娜, 刘莹, 党皓飞, 等. 2014.集合数值预报在洪水预报中的应用进展[J].应用气象学报, 25(6):641-653.
