基于PSRT与Elman神经网络的短期负荷多步预测  

Short-term Load Multi-step Forecasting Based on PSRT and Elman Neural Network

作　　者：李如琦[1] 孙艳[1] 孙志媛[2] 唐卓贞[1] 

机构地区：[1]广西大学电气工程学院,南宁530004 [2]广西电力试验研究院有限公司,南宁530023

出　　处：《电力系统及其自动化学报》2007年第6期84-87,113,共5页Proceedings of the CSU-EPSA

摘　　要：短期负荷预测(short-term load forcastings,STLF)对电力系统的经济和安全运行有着重要的作用。为提高短期负荷预测的精度,根据短期负荷的基本特性,提出了一种将相空间重构理论(phase space reconstruction space,PSRT)与Elman神经网络相结合的短期负荷多步预测模型。首先利用PSRT重构相空间的吸引子,然后用Elman神经网络来拟合相空间吸引子的多步演化,其中利用空间欧氏距离来选取Elman网络的输入样本。通过对广西电网短期负荷预测的分析表明,该多步预测方法是有效可行的。Short-term load forecasting （STLF） is important to the economic and secure operation of power systems. In order to improve STLF accuracy, a multi-step forecasting model of STLF based on phase space reconstruction theroy（PSRT） and Elman neural networks is presented in this paper according to essential features of power short-term load. Firstly, attractors in phase spaces are reconstructed by using PSRT. Secondly, the attractor＇s multi-step evolvement is fitted by using Elman neural networks and then the networks＇ input data of Elman network are selected using space Euclid distance. The results analysis of Guangxi grid＇s STLF shows that the proposed multi-step forecasting method is feasible and effective.

关 键 词：相空间重构 混沌时间序列 短期负荷预测 ELMAN神经网络 欧氏距离 

分 类 号：TM715[电气工程—电力系统及自动化]