96点短期负荷预测方法及其应用  

NINETY- SIX POINTSSHORT- TERM LOAD FORECASTING—— THEORY &APPLICATIONS

作　　者：鞠平[1] 姜巍 赵夏阳 王俊锴 张世学 刘琰 

机构地区：[1]河海大学电力系,南京210098 [2]陕西省电力公司,西安710004

出　　处：《电力系统自动化》2001年第22期32-36,共5页Automation of Electric Power Systems

基　　金：教育部高校优秀青年教师奖励计划资助项目

摘　　要：结合相似日方法和人工神经网络方法的优点 ,提出一种基于相似日和径向基函数网络的新方法。该方法具有较好的稳定性和精度 ,与普通相似日方法相比 ,减少了对人的经验的依赖 ,但仍然保持了应用简便、实用有效的特点。开发了 96点短期负荷预测软件 ,并在陕西电网成功运行 1年多。给出了几种方法的实际应用结果 ,并进行了比较 ,统计表明新方法的平均均方根误差较普通相似日方法从 3.1 5 3%降低到 2 .95 4 % ,用综合方法则进一步降低到 2 .85 4 %。Based on similar day method and radial basis function network, this paper proposes a new method for short-term load forecasting. The new method uses the parameters of several more similar days, instead of only one similar day in similar day method, as the inputs to radial basis function network for forecasting the loads of 96 points (15 minutes per point) of the next day. The method behaves the advantages of both similar day method and radial basis function network method, i.e. simple, practical, accurate and experience unreliable. The software developed has run well for over one year in Shanxi Provincial Power Company. The practical application results show that the mean error reduces from 3.153% to 2.954% compared with conventional similar day method.

关 键 词：电力系统 相似日方法 径向基函数网络 短期负荷预测 人工神经网络 

分 类 号：TM714[电气工程—电力系统及自动化]