短期风速的Adaboost＿GRNN组合预测模型  

Adaboost＿GRNN Combination Forecasting Model for Short-term Wind Speed

作　　者：芦婧 曾明 LU Jing;ZENG Ming(Department of Automation,Taiyuan Institute of Technology,Taiyuan 030008,China;School of Electrical and Information Engineering,Tianjin University,Tianjin 300072,China)

机构地区：[1]太原工业学院自动化系,太原030008 [2]天津大学电气自动化与信息工程学院,天津300072

出　　处：《电力系统及其自动化学报》2019年第4期70-76,共7页Proceedings of the CSU-EPSA

基　　金：国家自然科学基金资助项目(61271321,61573253);天津市科技支撑计划重点资助项目(14ZCZDS F00025);太原工业学院科学基金资助项目(2018LG04)

摘　　要：对风电场风速进行准确预测对于风能的开发利用具有重要意义。为了克服单一预测方法的局限性并进一步提高预测精度,提出了基于Adaboost算法和广义回归神经网络的短期风速组合预测方法。首先,分别采用时间序列法、支持向量机法和神经网络法建立3种风速预测模型;其次,采用广义回归神经网络将这3种单一模型的预测值进行非线性组合;最后,利用Adaboost算法集成多个广义回归神经网络的输出并将其作为高精度的风速预测值。算例测试结果表明,所提组合方法的预测精度高于各个单一模型以及熵权法组合模型和广义回归神经网络组合模型的预测精度。An accurate forecasting of wind speed in wind farms is significant for the exploitation and utilization of wind energy. To overcome the limitations of single forecasting models and further improve the forecasting accuracy,a combination forecasting method for short-term wind speed is proposed based on the Adaboost algorithm and general regression neural network(GRNN). Firstly,three wind speed forecasting models are established by means of time series,support vector machine(SVG)and neural network methods,respectively. Secondly,the predictions of these three single forecasting models are combined nonlinearly by adopting GRNN. Finally,the outputs from multiple GRNNs are combined using the Adaboost algorithm and further taken as high-precision wind speed predictions. The test result of a case study shows that the accuracy of the proposed combination forecasting method is higher than those of all single models and those of combination models based on the entropy weight method and GRNN,respectively.

关 键 词：风电场 短期风速预测 ADABOOST算法 广义回归神经网络 组合预测模型 

分 类 号：TM614[电气工程—电力系统及自动化]