基于小波分解的DIF-RBFNN超短期风速组合预测方法  

Ultrashort-term combined forecasting method of wind speed based on DIF-RBFNN with wavelet decomposition

作　　者：李德顺[1,2,3] 李宁 李银然[1,2,3] 吴世龙 李仁年[1,2,3] 郭涛 LI De-shun;LI Ning;LI Yin-ran;WU Shi-long;LI Ren-nian;GUO Tao(College of Energy and Power Engineering,Lanzhou Univ.of Tech.,Lanzhou 730050,China;Wind Turbine Engineering Technology Research Center of Gansu Province,Lanzhou Univ.of Tech.,Lanzhou 730050,China;Key Laboratory of Fluid Machinery and System of Gansu Province,Lanzhou Univ.of Tech.,Lanzhou 730050,China;Ningxia Jiaze New Energy Co.Ltd.,Yinchuan 750001,China)

机构地区：[1]兰州理工大学能源与动力工程学院,甘肃兰州730050 [2]兰州理工大学甘肃省风力机工程技术研究中心,甘肃兰州730050 [3]兰州理工大学甘肃省流体机械及系统重点实验室,甘肃兰州730050 [4]宁夏嘉泽新能源股份有限公司,宁夏银川750001

出　　处：《兰州理工大学学报》2019年第4期63-66,共4页Journal of Lanzhou University of Technology

基　　金：国家自然科学基金(51166009,51566011);国家重点基础研究发展计划(973计划)(2014CB046201);国家高技术研究发展计划(863计划)(2012AA052900);甘肃省自然科学基金(145RJZA059)

摘　　要：提出了一种基于小波分解(wavelet-decomposition)的数据输入格式-径向基神经网络(data input format-radial basis functional neural network)超短期风速组合预测模型.该模型首先将风速时间序列数据进行小波分解,减缓风速时间序列的波动性,然后将分解后的低频、高频部分分别建立数据输入格式(风速输入矩阵),并通过径向基神经网络模型进行预测,最后通过自适应叠加得到最终预测结果.结合宁夏某风场实测数据,将该预测模型和其他三种预测模型的仿真实验结果与实测值进行对比,表明该组合预测模型具有较高的预测精度.A new combined model of ultrashort-term forecast of wind speed was presented based on data input format radial basis functional neural network with wavelet decomposition.In this model,the date of original wind speed time series was decomposed with wavelets first to alleviate the fluctuation of wind speed time series.Then the data input fermat of the wind speed (wind speed input matrix) was set up respectively for the decomposed portions with low and high frequency and the wind speed was predicted by means of radial basis functional neural network model,and finally,the ultimate forecasting result was acquired with adaptive superposition.By means of actually measured data of a wind farm in Ningxia,the result of this prediction model was compared with that of simulative test and actual measurement of other three prediction models and it was shown that this combined forecasting model would have a higher accuracy of wind speed forecasting.

关 键 词：数据输入格式 小波分解 径向基神经网络 

分 类 号：TK83[动力工程及工程热物理—流体机械及工程]