基于人工神经网络的风电功率短期预测研究  

Study of short-term wind power prediction based on artificial neural networks

作　　者：彭怀午[1] 刘方锐[2] 杨晓峰[1] 

机构地区：[1]内蒙古电力勘测设计院,呼和浩特010020 [2]华中科技大学电气与电子工程学院,武汉430074

出　　处：《华东电力》2009年第11期1918-1921,共4页East China Electric Power

基　　金：国家自然科学基金重点项目(50837003);国家重点基础研究发展计划项目(2009CB219701)

摘　　要：采用人工神经网络方法对内蒙古某风电场10台风机的输出功率进行了预测研究。分别利用不同输入变量和网络结构的4种网络模型进行了预测分析,发现采用风速、风向和温度作为输入变量,4-8-1型的网络结构,预测精度较高,平均绝对误差小于250 kW,正规化均方根误差小于5%。The artificial neural network based wind power prediction method was employed to predict the output power of 10 wind turbines of a wind farm in Inner Mongolia.Four models with different inputs and net structures were used to perform the wind power prediction.It was found that the model with the 4-8-1 net structure and using wind speeds,wind directions,and temperature as inputs exhibits the best prediction accuracy with the mean absolute error less than 250 kW and the normalized root mean square error less than 5 percent.

关 键 词：风力发电 风电短期功率预测 人工神经网络 

分 类 号：TK81[动力工程及工程热物理—流体机械及工程]