基于偏最小二乘回归分析的中长期电力负荷预测.pdfVIP

第 32 卷 第 19 期 电 网 技 术 Vol. 32 No. 19 2008 年 10 月 Power System Technology Oct. 2008 文章编号：1000-3673 （2008 ）19-0071-07 中图分类号：TM715 文献标志码：A 学科代码：470·4054 基于偏最小二乘回归分析的中长期电力负荷预测 毛李帆，江岳春，龙瑞华，李妮，黄慧，黄珊 （湖南大学电气与信息工程学院，湖南省 长沙市 410082 ） Medium- and Long-Term Load Forecasting Based on Partial Least Squares Regression Analysis MAO Li-fan ，JIANG Yue-chun ，LONG Rui-hua ，LI Ni ，HUANG Hui ，HUANG Shan （College of Electrical and Information Engineering，Hunan University ，Changsha 410082，Hunan Province ，China ） ABSTRACT: Aiming at medium- and long-term load 接影响电网规划的质量。最小二乘回归分析方法是 forecasting, the principle of partial least squares regression 负荷预测中较常用的方法，算法简单、运算量小是 analysis is presented and the simplified modeling procedures of 该方法的一大优点。然而，当自变量系统中存在严 this algorithm is derived in detail. Under the presupposition of 重的多重相关性时，最小二乘法拟合出的回归模型 keeping original data information as possible, the proposed 的精确性、可靠性都不能得到保证；并且在样本个 method concentrates the data information to several principal 数与自变量个数相比较少的情况下，不得不采用向 components uncorrelated each other, thus the difficulty of not enough sample numbers and severe multiple correlation of 前选择变量法、向后删除变量法或逐步回归法等方 self-variables that makes it hard to build forecasting model by 法对自变量进行筛选，这无疑增加了负荷预测工作 multiple regression analysis can be effectively solved while the 的负担[1] 。 load forecasting model is built. By means of calculation 偏最小二乘回归分析是一种新型的多元统计 example, the partial least squares regression analysis method, 数据分析方法，它集多元线性回归分析、典型相关 the least squares method and the stepwise regression analysis 分析和主成分分