基于方差寻优定权的电量预测方法
Electricity Consumption Prediction with Variance-based Optimum Searching Weighting
王洋 1李江 1张婧 2格日乐图 3修子孟4
作者信息
- 1. 内蒙古电力(集团)有限责任公司,内蒙古呼和浩特 010000
- 2. 内蒙古电力(集团)有限责任公司数字研究院分公司,内蒙古呼和浩特 010020
- 3. 内蒙古电力集团蒙电信息通信产业有限责任公司,内蒙古呼和浩特 010020
- 4. 北京清软创新科技股份有限公司,北京 100080
- 折叠
摘要
近年来气候变化的影响和用电量变化的不规律性,对月度电量预测提出了更高要求.单一的预测方法无法取得理想的预测效果,因此需要考虑多种预测方法的寻优加权以提升预测精度.基于对历史数据规律性的分析,寻找适于发展规律的多种模型进行预测,开展方差寻优定权方法研究,得到综合预测结果.首先,提取某省农业指标历史电量,使用多种算法预测;其次,将各预测值分别融入历史数据进行同比、环比方差测算,对同比、环比维度序列方差平均值按波动程度自动寻优分配权重,最终加权得到预测结果;最后,应用该方法进行预测,验证了其有效,且具有较高预测精度.
Abstract
In recent years,with the impact of climate change and the irregularity of electricity consumption changes,high-er requirements have been put forward for monthly electricity prediction.Individual prediction methods cannot achieve ideal prediction results,so it is necessary to consider the optimization weighting through hybrid prediction methods to im-prove accuracy.By analyzing and exploring the regularity of historical data,this work sought various models suitable for predicting development patterns,conducted methodological study on variance-based optimum searching and weighting,and obtained comprehensive prediction results.First the historical electricity consumption of agricultural indicators in a certain province was extracted and multiple algorithms were selected for prediction.Second all the independent predicted values were integrated into historical data for year-on-year and month-on-month variance calculation.The average variance of both of the sequence variance values were automatically weighted according to the degree of fluctuation,and the final weighted prediction result was obtained.The proposed method was applied for practical prediction,which verified its ef-fectiveness and high prediction accuracy.
关键词
方差寻稳/自动分配权重/EVIEWS/VAR向量自回归/SARIMAKey words
variance-based optimum searching/automatic weighting/EVIEWS/VAR vector autoregression/SARIMA引用本文复制引用
出版年
2024
NETL
NSTL
万方数据