电力短期负荷的多嵌入维一阶局域预测.pdfVIP

29 卷 13 期 电 网 技 术 Vol. 29 No. 13 2005 年7 月 Power System Technology Jul. 2005 文章编号：1000-3673 （2005 ）13-0045-05 中图分类号：TM715 文献标识码：A 学科代码：470 4051 电力短期负荷的多嵌入维一阶局域预测 雷绍兰，孙才新，周 湶，张晓星 （高电压与电工新技术教育部重点实验室（重庆大学），重庆市 沙坪坝区 400030 ） SHORT-TERM LOAD FORECASTING USING ONE-RANK LOCAL-REGION METHOD IN MULTI-DIMENSION EMBEDDING PHASE SPACE LEI Shao-lan，SUN Cai-xin，ZHOU Quan，ZHANG Xiao-xing （The Key Laboratory of High Voltage Engineering and Electrical New Technology of Ministry of Education （Chongqing University ），Shapingba District，Chongqing 400030 ，China ） ABSTRACT: In chaotic theory there is a litter difference in the [1,2] [3-6] 传统分 法 和现代智能法 。传统方法通常仅 obtained dimension by different embedded dimension 以历史负荷作为预测的主要依据，其计算方法简 calculation methods, and under different embedded dimension 单，但预测精度难以令人满意；现代智能技术在灵 the forecasting results for same time series of load are different 活选择输入样本和智能学习方面较传统方法均有 too. For this reason, the authors propose a load forecasting 较大改进，但由于影响电力负荷的因素很多，在实 method using multi-embedded dimension, i.e., applying the weighted average to forecasting results under different 际预测中很难将所有因素都考虑进去，这一方面是 dimension. Because the values of Euclid distance limits used in 由于收集某些历史数据困难，另一方面是 考虑因 the forecasting evidently impact the forecasting results, so in 素太多时网络结构过于复杂难以实现。 this paper the value of Euclid distance limit is chosen by 近年来，混沌理论引起了许多学者的普遍关 dynamic adj ustment method to make the forecasting erro