CAJ | 학술논문

为了解决风电功率神经网络预测输入变量多、计算效率低、泛化能力较差的缺点,采用主成分分析法(PCA)减少变量数.用神经网络动态集成的方法构建出较强泛化能力的BP网络集成.采用南方某风电场的数据进行了预测,比较了选取全部气象参数、部分气象参数和基于PCA处理后的数据作为神经网络输入对预测精度和计算效率的影响,结果表明采用PCA能在不降低预测精度的情况下,大大提高运算速度.通过对比单个和集成BP神经网络预测结果发现,采用集成网络的预测精度比单个BP网络精度有所提高,特别是风速突变的情况下更加明显.
위료해결풍전공솔신경망락예측수입변량다、계산효솔저、범화능력교차적결점,채용주성분분석법(PCA)감소변량수.용신경망락동태집성적방법구건출교강범화능력적BP망락집성.채용남방모풍전장적수거진행료예측,비교료선취전부기상삼수、부분기상삼수화기우PCA처리후적수거작위신경망락수입대예측정도화계산효솔적영향,결과표명채용PCA능재불강저예측정도적정황하,대대제고운산속도.통과대비단개화집성BP신경망락예측결과발현,채용집성망락적예측정도비단개BP망락정도유소제고,특별시풍속돌변적정황하경가명현.
The wind power artificial neural network (ANN) forecasting has shortcomings such as a large amount of variables, low computation efficiency and poor generalization ability. This paper proposes to apply the principal components analysis (PCA) to reduce the number of variables. Neural network dynamic integrating is adopted to establish the BP network integration with stronger generalization ability. Data of a wind power station in the South is used to forecast and compare the influence on accuracy and computation efficiency exerted by neural network input of all meteorological parameters, part of meteorological parameters and data based on PCA processing respectively. It is shown that using PCA processing can improve the computation efficiency greatly while keeping the forecast accuracy. Comparing the neural network forecasting results of single BP net with those of integrating BP net, we found that the latter one can perform better in improving the accuracy, especially in the case of sudden change of wind speed.