基于偏最小二乘回归与比重法的月售电量预测  

Forecasting for Monthly Electricity Consumption Using Partial Least-square Regressive and Proportion Model

作　　者：吴杰[1] 宋国堂[1] 卢志刚[1] 张鸿 

机构地区：[1]燕山大学电气工程学院,秦皇岛066004 [2]唐山电力公司,唐山063000

出　　处：《电力系统及其自动化学报》2008年第3期66-69,共4页Proceedings of the CSU-EPSA

摘　　要：月售电量的预测受多方面的制约,从影响售电量的因素出发,利用偏最小二乘回归与比重法建立了国民生产总值、人口、社会固定资产投资、人均国民生产总值与售电量的回归预测模型。偏最小二乘方法能够提取若干对系统具有最佳解释能力的综合变量来建立预测模型[1],该方法与比重法结合应用于月售电量的预测之中,能更好地体现引起月售电量变化的平稳因素、季节突变因素的周期性,使得月售电量的预测更加准确。利用该预测模型对唐山地区2004年的月售电量进行了预测,月售电量的平均相对误差为4.74%,预测精度较高,证明了该预测模型的准确性。Electricity consumption is affected by many factors,including gross national product,population, social fixed assets and so on. This paper proposes a forecasting model comprehensively considering the relationship between the factors above and electricity consumption based on partial least-square regressive and proportion method. Partial least-square regressive can give the most expositive elements to establish the forecasting model. Combined with proportion method, it can reflect the seasonal change factors causing the variation of electricity consumption, which makes the forecasting of monthly electricity consumption more accurate. The model has been applied to forecast 2004 monthly electricity consumption in Tangshan area. With 4. 74% average relative error ,the results show the validity of the model.

关 键 词：偏最小二乘回归 比重法 月售电量预测 多元线性回归 

分 类 号：TM715[电气工程—电力系统及自动化]