PCA-RBF神经网络模型在工业用户电力负荷预测中的应用  

Application of PCA-RBF Neural Network Model in Industrial User's Power Load Forecasting

作　　者：田娜[1] 鞠黄培 王旭东[1] 黄莉[3] 郑红娟[3] 

机构地区：[1]国网天津市电力公司,天津300384 [2]天津大学,天津300072 [3]国电南瑞科技股份有限公司,江苏南京211106

出　　处：《电器与能效管理技术》2016年第21期58-62,70,共6页Electrical & Energy Management Technology

基　　金：国家重点专项基金(2016YFB0901103);国网科技项目(SGTJ0000KXJS1400087)

摘　　要：依据工业用户生产性负荷特征,运用主成分分析法(PCA)对工业用户的生产性负荷因子进行标准化处理,评估特征值、特征向量及累计方差贡献率等指标,获得携带工业用户主要影响因素的关键特征量;结合径向基函数(RBF)神经网络算法,构建基于主成分分析(PCA)和RBF神经网络组合(PCA-RBF)的生产性负荷预测模型。试验结果表明,PCA-RBF神经网络预测模型能有效克服传统神经网络训练速度慢的不足,且具有更高的预测精度。The nonlinear and highly redundancy relationship of productive factors for studying load prediction of industries increases difficulty and accuracy of prediction. A method based on principal component analysis（ PCA） and radius basis function（ RBF） neural network was proposed to predict the process load. As for industries with highdimensional process load factors,series of steps are taken,including normalized-process,correlation matrix establishment,eigenvalues,eigenvectors and cumulative variance contribution rate calculation. Combined effect factors carrying most useful information were chosen to be inputs of RBF neural network process load prediction model,and the process load is the output of the model. One typical industry process load data were tested and validated,and the simulation results demonstrate that the proposed PCA-RBF neural network load prediction model can effectively overcome the deficiency of training rate and has higher prediction accuracy than traditional neural network.

关 键 词：电力负荷 主成分分析法 神经网络 预测模型 

分 类 号：TM734[电气工程—电力系统及自动化]