基于季节调整和BP神经网络的月度负荷预测  

Monthly load forecasting model based on seasonal adjustment and BP neural network

作　　者：龙勇[1] 苏振宇[1,2] 汪於[1] LONG Yong;SU Zhenyu;WANG Yu(School of Economics and Business Administration, Chongqing University, Chongqing 400030, China;Training Center, Gansu Electric Power Corporation, Lanzhou 730070, China)

机构地区：[1]重庆大学经济与工商管理学院,重庆400030 [2]甘肃省电力公司培训中心,兰州730070

出　　处：《系统工程理论与实践》2018年第4期1052-1060,共9页Systems Engineering-Theory & Practice

基　　金：国家社会科学基金重点项目(14AZD130)~~

摘　　要：月度电力负荷序列中离群值及节假日因素会影响月度负荷预测的准确性．为此，提出了基于季节调整方法和BP神经网络的月度电力负荷组合预测模型．首先，利用季节调整方法对原始负荷序列进行预处理，消除离群值和春节假日的影响；然后用BP神经网络对回归残差序列建模预测得到预测结果或对季节调整后序列和季节成分序列分别建模预测，并对分量预测结果重构后得到最终预测结果的方法．通过实例对预测效果进行检验，结果表明提出的预测方法的预测表现要优于BP神经网络，SARIMA，支持向量机等模型，可以获得更高的预测精度．The accuracy of monthly load forecasting will be affect by the outliers and holidays such as Spring Festival. So, two kinds of monthly load forecasting model based on seasonal adjustment and BP neural network was put forward. One, the original load series was pre-adjusted by using the seasonal adjustment method, the impact of outliers and the Spring Festival effect was eliminated; then BP neural network was used on the regression residuals series, the forecasting results can be attained. Another, the original load series was pre-adjusted as same as above, then the BP network was used on seasonally adjusted series and seasonal component series separately, the final forecasting results attained through reconstruct the forecasting result of component series. Case study results show that the prediction accuracy of proposed methods was better than BP neural network, SARIMA, support vector machines and other methods.

关 键 词：负荷预测 季节调整 离群值 神经网络 

分 类 号：TM715[电气工程—电力系统及自动化]