一种PLS回归的并网风电项目利润预测方法  

A profit forecast method of wind power project based on partial least squares regression

作　　者：张戈力[1] 毛安家[1] 赵岩 

机构地区：[1]华北电力大学电气与电子工程学院,北京102206 [2]华电国际电力股份有限公司,北京100031

出　　处：《电力系统保护与控制》2013年第8期87-92,共6页Power System Protection and Control

基　　金：中央高校基本科研业务专项资金支持项目(12MS20)

摘　　要：为了减少风电项目的盲目建设,提高建设投资决策的经济性,提出一种基于偏最小二乘回归的风电项目利润总量预测方法。利用偏最小二乘回归算法中的指标标准化处理和提取偏最小二乘成分,解决了实际利润预测分析中在计算数据的指标之间具有多重相关性、可以搜集到的准确的相关样本数据较少的条件下拟合利润预测回归模型时存在的问题。提出影响风电项目年利润总量的九个指标,在某地若干风电项目的实例数据的基础上,计算了实例数据的膨胀因子并进行特异点分析,利用偏最小二乘法建立了利润预测模型,探讨了其与传统的回归预测方法的差异。算例分析结果表明,该方法在预测项目年利润时预测精度较高,是一种有效可行的预测方法。In order to reduce blind wind power construction and to improve construction economy, a wind power project profit forecasting model based on PLS method is proposed. Data standardization method and PLS principal components are used to solve the issue of multiple correlations between data and the problem of inadequate of original sample data which would cause regression fitting problems in profit prediction. The principle of partial least squares regression analysis is firstly introduced, and nine indexes influencing the annual profit are proposed. Based on an actual data VIF calculation and sample outliers analysis are performed. A profit prediction model is built based on the PLS method. Finally, the difference between the PLS analysis method and normal regression methods is also discussed. The result proves the accuracy of the model and practicable utility of the method.

关 键 词：风电项目 利润预测 多重相关 偏最小二乘回归法 

分 类 号：TM614[电气工程—电力系统及自动化]