本文選題：核主成分分析 + 核最小最大概率回歸機(jī)�。� 參考：《電力自動(dòng)化設(shè)備》2017年02期

【摘要】：針對(duì)短期風(fēng)電功率概率預(yù)測(cè),提出一種基于核主成分分析(KPCA)與核最小最大概率回歸機(jī)(KMPMR)相結(jié)合的方法。KPCA方法可對(duì)數(shù)據(jù)進(jìn)行預(yù)處理,在特征空間中有效提取模型輸入的非線性主元;KMPMR方法在僅需假定產(chǎn)生預(yù)測(cè)模型的數(shù)據(jù)分布的均值與協(xié)方差矩陣已知時(shí),將最小最大概率分類機(jī)(KMPMC)的分類超平面看作預(yù)測(cè)模型的輸出,可最大化模型的輸出位于其真實(shí)值邊界內(nèi)的最小概率。實(shí)驗(yàn)結(jié)果表明,所提方法在預(yù)測(cè)精度上優(yōu)于現(xiàn)有的預(yù)測(cè)方法,并能提供預(yù)測(cè)誤差的分布范圍。
[Abstract]:For short-term wind power probability prediction, a method based on kernel principal component analysis (KPCA) and kernel minimum and maximum probability regression (KMPMRs) is proposed. KPCA can preprocess the data. The nonlinear principal component KMPMR method, which effectively extracts the input from the model in the feature space, only needs to assume that the mean value and covariance matrix of the data distribution generated by the prediction model are known. The classification hyperplane of the minimum maximum probability classifier (KMPMC) is regarded as the output of the prediction model, which maximizes the minimum probability that the output of the model lies within its real value boundary. The experimental results show that the proposed method is superior to the existing prediction methods in prediction accuracy and can provide the distribution range of prediction errors.
【作者單位】： 蘭州交通大學(xué)自動(dòng)化與電氣工程學(xué)院;
【基金】：國(guó)家自然科學(xué)基金資助項(xiàng)目(51467008)~~
【分類號(hào)】：TM614
，

本文編號(hào)：1994675