学术
径流混沌时间序列局域多步预测模型及其在黄河上游的应用
张文鸽[1,2] 黄强[1] 佟春生[3]
[1]西安理工大学水利水电学院,西安710048 [2]黄河水利科学研究院水资源研究所,郑州450003 [3]华北工学院分院,太原030008
摘 要:
近10多年来，许多学者相继开展了应用混沌理论对径流时间序列的预测研究，以Takens嵌入定理为理论基础的混沌局域法是一种简单、有效的预测方法。但是常用的零阶局域法、一阶局域法、加权零阶局域法和加权一阶局域法都是一种单步预测模型，进行多步预测时计算量大且存在误差累积效应。基于相空间重构技术的加权一阶局域法多步预测模型可以克服上述不足。因此，本文首先利用虚假邻域法选取相空间重构的参数时间延迟和嵌入维数，而后依据小数据量法计算最大Lyapnuov指数进行径流时间序列混沌特性的定量识别，最后建立了径流混沌时间序列加权一阶局域法多步预测模型，并将该模型应用于黄河上游贵德站1954年1月-2003年12月的实测径流时间序列预测。结果表明，该模型用于径流时间序列的短期预测是可行和有效的。[著者文摘]
关 键 词:
文章出处:
《水力发电学报》-2007年26卷4期 -11-15页
分 类 号:
文献标识码:
A
Chaotic local-region multi-step and its application on the forecasting model of flow time series upper reach of the Yellow river
ZHANG Wenge, HUANG Qiang, TONG Chunsheng （ 1. School of Water Conservancy and hydropower, Xi＇ an University of Technology, Xi＇ an 710048 ; 2. Water Resources Research Institute of Yellow River Hydraulic Research Academy, Zhengzhou 450003 ; 3. North China College of Technology, Taiyuan 030008）
Abstract:
Chaos predictions of flow time series are studied by many scholars in the past ten years. Local-region method based on Takens theory is a simple and valid one. But zero-rank, one-rank, adding-weight zero-rank and one-rank local-region methods are all single-step predictions. There are large computations and cumulative errors when multi-step predictions are carried out. An adding-weight one-rank local-region multi-step forecasting model based on phase reconstruction can overcome the shortcomings. Firstly. phase space reconstruction parameters of time delay and embedment dimension are chosen by false nearest neighbor method; Then, chaotic characteristic of flow time series is identified by computing the largest Lyapunov index. Finally, the model of flow time series prediction by using adding-weight one-rank local-region method is established and Gui De monthly flow time series prediction from January to December on the upper reach of the Yellow River is studied. The results show that the model for short-term flow time series prediction is valid.[著者文摘]
Key words:
chaotic identification ; time series ; flow prediction ; phase-space reconstruction ; largest lyapunov undex
基金资助:
国家自然科学基金项目（E090350239090）;河南省自然科学基金项目（0511053300）
更多>>免费文章
cqvip.com