采用最优子集回归模型的风电场风速预测方法  

Prediction method of wind speed on wind farm using optimal sub-set regression model

作　　者：窦震海[1,2] 杨仁刚[1] 邹黎[2] 唐诗[2] 于茂荣[3] 姜李[2] 

机构地区：[1]中国农业大学信息与电气工程学院,北京100083 [2]山东理工大学电气与电子工程学院,山东淄博255049 [3]山东理工大学商学院,山东淄博255049

出　　处：《可再生能源》2014年第8期1161-1167,共7页Renewable Energy Resources

基　　金：国家高技术研究发展计划("863"计划)项目(2012AA050217)

摘　　要：风速的快速准确预报是电力系统短期调度计划的基础,为了进一步提高风速预测的精度和运算速度,针对短期风速样本数据兼具波动性和趋势性的特点,采用均生函数最优子集回归(MGF-OSR)方法建立预测模型。该模型引入了一阶、二阶差分序列拟合序列的高频部分,建立累加生成序列来拟合序列的趋势性,利用均值生成函数(MGF),对上述所有序列创建预测因子矩阵,根据双评分准则(CSC)进行粗选,剔除评分较低的因子,对其他预报因子进行寻优组合得到最优子集,并以此建立预测模型。实例分析表明,该模型预测的平均相对误差大大低于其他预测方法,预测精度优势明显。Quick and accurate forecast of wind speed is the basis of power system short-term scheduling. To improve the precision and speed of wind speed forecasting, according to the trend and fluctuation of short-term wind speed forecasting sample data,a short-term wind speed forecasting method of optimal subset regression based on mean generating function （MGF-OSR） is proposed in this paper. This method establishes a prediction model by MGF-OSR. This method introduces first-order and second-order difference sequence to fit the high frequency part, and establishes accumulation generating sequence to fit the trend. All of the above sequences are used to form prediction factors matrix by MGF. This method adopted couple score criterion （CSC） for roughly excluding lower score factors. The optimal subset can be got by combination and optimization of the remaining forecasting factors, and then the forecasting model is built. Comparing with common methods, by analysis of case study, the results show that this forecasting model＇s average relative error can be reduced obviously, the prediction accuracy of MGF-OSR is obviously better than other methods.

关 键 词：短期风速预测 最优子集 双评分准则 

分 类 号：TK81[动力工程及工程热物理—流体机械及工程]