基于改进GM(1,1)和SVM的长期电量优化组合预测模型  

An optimally combined forecast model for long-term power demand based on improved grey and SVM model

作　　者：宋晓华[1] 祖丕娥[1] 伊静[1] 刘达[1] 

机构地区：[1]华北电力大学经济与管理学院,北京102206

出　　处：《中南大学学报（自然科学版）》2012年第5期1803-1807,共5页Journal of Central South University:Science and Technology

基　　金：国家自然科学基金资助项目(70901025);中央高校基本科研业务费专项资金资助项目(11MR22)

摘　　要：针对中长期电量预测可使用的相关历史数据较少、影响因素较为复杂等特点,提出一种基于改进GM(1,1)和支持向量机的优化组合预测模型。该模型将改进灰色预测模型和支持向量机模型进行组合,采用蛙跳寻优算法求取组合预测模型中各单一模型的权重,构建基于蛙跳优化的组合预测模型。将优化后的组合预测模型应用于我国中长期电量预测,选择我国1991—2005年电量进行分析,对2006—2010年的电量进行预测,并与一般组合预测模型及各单一模型进行比较。研究结果表明:本文方法得到的电量平均相对误差为2.06%,比等权组合预测模型、方差-协方差优选组合预测模型以及各单一预测模型的预测精度都有所提高。According to the characteristics of long-term power demand forecasting,an optimally combined forecast model based on improved GM（1,1） and support vector machine was proposed.In this model,the shuffled frog leaping algorithm was used to optimize the weight of each model.The optimized combination forecast model was used to forecast the long-term power demand of our country,based on the training exampling power demand from 1991 to 2005,the testing exampling power demand from 2006 to 2010,and it was compared with common combination forecast model and each single model.The results show that the average relative error is 2.06% using the combined forecast model.Compared with equal weight combined forecast model,variance-covariance optimally combined forecast model and the single model,the prediction accuracy is improved.

关 键 词：组合预测 蛙跳算法 灰色预测 支持向量机 

分 类 号：TU457[建筑科学—岩土工程] TU413.6[建筑科学—土工工程]